The Making of a Mixed Language: The Case of Ma'a Mbugu (Creole Language Library)

The Making of a Mixed Language

Creole Language Library (CLL) A companion series to the “Journal of Pidgin & Creole Languages
” Editors Jacques Arends

John Victor Singler

Amsterdam

New York

Editorial Advisory Board Mervyn Alleyne

Salikoko Mufwene

Kingston, Jamaica

Chicago

Norbert Boretzky

Pieter Muysken

Bochum

Nijmegen

Lawrence Carrington

Peter Mühlhäusler

Trinidad

Adelaide

Glenn Gilbert

Pieter Seuren

Carbondale, Illinois

Nijmegen

George Huttar

Norval Smith

Dallas

Amsterdam

John Holm Coimbra

Volume 26 The Making of a Mixed Language: The case of Ma’a/Mbugu by Maarten Mous

The Making of a Mixed Language The case of Ma’a/Mbugu

Maarten Mous Leiden University

John Benjamins Publishing Company Amsterdam
/
Philadelphia

8

TM

The paper used in this publication meets the minimum requirements of American National Standard for Information Sciences – Permanence of Paper for Printed Library Materials, ansi z39.48-1984.

Library of Congress Cataloging-in-Publication Data Maarten Mous The Making of a Mixed Language : The case of Ma’a/Mbugu / Maarten Mous. p. cm. (Creole Language Library, issn 0920–9026 ; v. 26) Includes bibliographical references and index. 1. Masai language--Grammar/ I. Title. II. Series. PL8501.1.M68 2003 496’.5-dc22 isbn 90 272 5248 3 (Eur.) / 1 58811 489 9 (US) (Hb; alk. paper)

2003063773

© 2004 – John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher. John Benjamins Publishing Co. · P.O. Box 36224 · 1020 me Amsterdam · The Netherlands John Benjamins North America · P.O. Box 27519 · Philadelphia pa 19118-0519 · usa

For Eithne

TABLE OF CONTENTS List of Tables List of Figures Acknowledgements List of Abbreviations Map 1 of the relevant area of east Africa Map 2 of the Usambara area 1 Introduction 1.1 The people and their language 1.2 Earlier descriptions and history of Ma'á studies 1.3 Source of data, fieldwork. 1.4 Areas of settlement of the Ma'á/Mbugu and clan affiliation 1.5 Multilingualism and language attitude 1.6 A sample of Mbugu language material 1.7 Properties of the parallel lexicon 1.8 New insights in this work 1.9 Organisation of this book 2 Historical and geographical background 2.1 Oral History 2.2 Historical information from the parallel lexicon 2.3 Usambara mountains: Shambaa, Seuta Bantu, Swahili 2.4 Maasai 2.5 Gorwaa and Mbugwe 2.6 Pare 2.7 The Taita connection 2.8 Old Kenyan Cushitic: Eastern Cushitic, Dahalo 2.9 Chronology through a study of lexical domains 2.10 Reconstructing the history: possible scenarios 3 Linguistic history 3.1 The issues and the debate 3.2 Absence of drastic recent linguistic developments 3.3 History of the lateral fricatives in Ma'á (Inner Mbugu) 3.4 Truncation rule 3.5 Remnants of non-Bantu grammar 3.6 Ma'á non-Bantu causatives

xi xiii xv xvi xviii xix 1 1 2 5 6 7 8 10 13 15 17 17 24 28 31 32 33 34 37 41 44 51 51 51 52 56 58 63

VIII 3.7 3.8 3.9 3.10 3.11 3.12 4 4.1 4.2 4.3 4.4 4.5 4.6 4.7 5 5.1 5.1.1 5.1.2 5.1.3 5.1.4 5.1.5 5.1.6 5.1.7 5.1.8 5.2 5.3 5.4 5.5 5.6 6 6.1 6.2 6.3 6.4 6.5 6.6 6.7 7 7.1 7.2 7.3 7.4

Non-Bantu nominal endings Locatives and other non-Bantu frozen grammatical elements The origin of function words Shift or gradual bantuisation Code-switching as a model for Ma'á Mixed languages as a challenge to historical linguistics Phonology Consonants Spirantisation in velars Prenasalised obstruents Vowels Tone Syllable structure Word structure and phonotactics The verb Overview Vowel coalescence, i-initial stems, and morphotonology Subject and object prefixes Other verbal prefixes, "tense" and polarity Relative marker The inflectional final vowel The a/e alternation Imperatives Clitics The "tenses" Combinations of "tense" prefixes The negative "tenses" Copula and verbs "to be" Verbal derivation The noun The noun class system The morphophonology of the noun class prefixes Noun class pairings Derivation Frozen suffixes Question nouns Names Adjectives and other nominal modifiers Adjectives Genetive and relative pronouns Possessives Demonstratives

68 71 72 74 S6 87 95 95 97 103 103 104 104 107 111 111 112 114 115 117 117 119 121 121 121 137 139 145 153 163 163 164 164 171 173 173 173 175 175 178 179 180

IX

7.5 8 8.1 8.2 8.3 8.4 9

Quantifiers Invariables or other words Dependent invariables: Prepositions Independent invariables Independent personal pronouns Higher numbers and other invariable modifiers Notes on syntax, code-switching and texts Syntax Set sentences Code-switching A sample of code-switching Texts

9.1 9.2 9.3 9.4 9.5 Notes References Appendix: Mbugu - English etymological lexicon English index to the lexicon Index

181 183 183 191 194 195 197 197 198 199 200 205 215 221 231 299 319

LIST OF TABLES Table 1.1: Area, clans, and language preference 7 Table 1.2: Derivation of Inner Mbugu by dummy suffix 13 Table 2.1: Oral traditions, clans 22 Table 2.2: Ma'á words that share innovations with Iraqw/Gorwaa 24 Table 2.3: Archaic Inner Mbugu words 25 Table 2.4: Diversity of origins of Inner Mbugu words 27 Table 2.5: Some Shambaa words in (Normal) Mbugu 29 Table 2.6: Inner Mbugu words of possibly Zigua origin 30 Table 2.7: Inner Mbugu words of Maasai origin 32 Table 2.8: Some lexical evidence for the South Pare affiliation of Normal Mbugu ... 33 Table 2.9: Inner Mbugu words shared with Taita (Sagala) 36 Table 2.10: Oromo cognates in Inner Mbugu 38 Table 2.11: Inner Mbugu words linked to Dahalo 39 Table 2.12: Yaaku - Inner Mbugu lexical resemblances 40 Table 2.13: Sample of fauna vocabulary and its origin 42 Table 2.14: Mbugu banana terminology with sources for the NMb terms 43 Table 2.15: Ma'á (Inner Mbugu) goat/sheep vocabulary and its origin 44 Table 3.1: The development of the lateral fricative i (hl) 53 Table 3.2: Surviving lateral fricatives 55 Table 3.3: Newly developed lateral fricatives 55 Table 3.4: Development of final é 57 Table 3.5: 1Mb causatives in -ti and their sources 64 Table 3.6: Ma'á words ending in ao 71 Table 3.7: Inner Mbugu demonstratives 72 Table 3.8: Maasai and Iraqw demonstratives 73 Table 3.9: Dahalo and Oromo demonstratives 73 Table 3.10:1Mb personal pronouns and their Cushitic equivalents 74 Table 3.11:1Mb possessives and their Cushitic equivalents 74 Table 3.12: Inner Mbugu Numerals and their origin 89 Table 4.1: Mbugu consonants 96 Table 4.2: (Near-)minimal pairs of l and r 97 Table 4.3: (Near-)minimal pairs of g and gh 98 Table 4.4: (Near-)minimal pairs  - x - h in Inner Mbugu 100 Table 4.5: (Dialect) variation between  - x - h in Inner Mbugu 101

XII Table 4.6: x>h in history Table 4.7: "Restoration" of x >  Table 4.8: Minimal pairs for tones Table 4.9: Possible tone patterns of tri-syllabic roots Table 4.10: Examples of syllable structure Table 4.11: Nasal + obstruent processes Table 4.12: Tone patterns for disyllabic noun roots Table 4.13: Tone patterns for mono-syllabic noun roots Table 4.14: Tone patterns for Normal Mbugu verb roots Table 4.15: Tone patterns for Inner Mbugu verb roots Table 5.1: Subject and object prefixes Table 5.2: "Tense" prefixes Table 5.3: Verbs with non-inflectional and inflectional final vowel a in 1Mb Table 5.4: Inflectional final vowels Table 5.5: The positive tenses Table 5.6: Structure of the examples for each "tense" in the list Table 5.7: Past tenses Table 5.8: Expressions of negation in verbs Table 5.9: Overview of verb forms with a negative pre-subject prefix te Table 5.10: Overview of verb forms with a negative "tense" prefix sV Table 5.11: The negative tenses with a negative pre-subject prefix te Table 5.12: Copula and verbs "to be" and their functions Table 5.13: Verbal derivational suffixes Table 5.14: The forms of the intensive/causative -ish Table 5.15: The forms of the causative -ij Table 5.16: Applied extension Table 5.17: Stative extension Table 5.18: The reciprocal Table 5.19: The extension -any Table 5.20: The passive extension Table 5.21: Applied-Passive extension Table 6.1: Mbugu noun classes (N refers to anasai element) Table 6.2: Singular/plural pairs of noun classes Table 6.3: Class pair 9/10 Table 6.4: Class 9a 0 , plural class 10a Ø Table 6.5: Class pair 11/10 Table 6.6: Class pair 11/6+11 Table 6.7: Noun class derivation Table 6.8: Nominalisations Table 7.1: All adjectives Table 7.2: Genitive and relative pronouns Table 7.3: Possessive stems

102 102 104 104 105 107 108 108 109 109 115 116 118 118 122 122 125 139 141 141 144 146 154 155 156 156 158 158 159 160 161 164 165 168 169 169 170 172 172 176 178 179

XIII Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table

7.4: Inner Mbugu demonstratives 7.5: Normal Mbugu demonstratives 7.6: Quantifiers 8.1: kwa in fixed expression meaning 'because, why' 8.2: Expressions with general locative use of cha 8.3: Relative days 8.4: Parts of the day 8.5: Locative expressions 8.6: Manner expressions 8.7: Conjunction-like invariables 8.8: Independent personal pronouns 8.9: The invariable numerals 8.10: Compound numbers 9.1: Sequence of greetings (1Mb) 9.2: Sequence of greetings (NMb)

180 181 182 187 190 191 192 193 194 194 195 195 196 198 198

LIST OF FIGURES Figure 1 : Word category and origin in Inner Mbugu Figure 2: The distribution over source languages for the Inner Mbugu lexicon

26 27

LIST OF ABBREVIATIONS 1/2 1PL 1SG

2 2SG

adv aj Al APL

As BE BGND

Bo Bu Bum C.EXPEC

C33 Ch COND COP CS CSEC D

Dl Dah Dav dem DEMI

E EVI EXPEC F FUT

noun class 1, plural 2 first person plural first person singular class 2 second person singular adverb adjective Alagwa applied Aasáx to be background Bowe = Mbugwe Burunge Bumbuli counter-expectational Copland 1933 Chaga conditional copula causative consecutive dependent demonstrative of first de­ gree Dahalo Davidą demonstrative demonstrative of first de­ gree Ehret (1980) evidential expectational (default) Final vowel future

G G58 Gw H HAB HERE HORT

I-4-NMb ideo 1Mb IMbA IMb-Bum IMb-Ra INC

infl.vowel INT

int inv Ir IT

L M06 Ma N n NARR NEG

NMb np num obj Or Pa

glide Green (1958) Gweno High tone habitual here hortative Inner and Normal Mbugu ideophone Inner Mbugu (Ma'á) Archaic Inner Mbugu Inner Mbugu Bumbuli dialect Inner Mbugu Rangwi dia­ lect incidental inflectional (final) vowel intensifier, causative interjection invariable Iraqw itive (thither) Low tone Meinhof(1906) Maasai nasal consonant noun narrative negative Normal Mbugu Noun Phrase numeral object Oromo Pare, Chasu.

ACKNOWLEDGEMENTS. Now that this book is finished it is time to thank the many people who were instru­ mental in its realisation. It all began with Thilo Schadeberg who first mentioned to me the need for a description of Ma'a and then convinced me that I could and should do it. Then Franz Rottland made the project possible by offering me a research position in Bayreuth which enabled me to do extensive fieldwork in Tanzania (SFB 214 Identität in Afrika A5). In Bayreuth I discussed many aspects of Ma'a with Derek Nurse and Mechthild Reh with whom I shared an office and I exchanged views with Abel Mreta when he was working there on his thesis on Pare (Chasu). My ideas about Ma'a were sharpened by the vigorous interest in my work from my office neighbour Gottfried Müller. I also profited a lot from the comments that I received on earlier versions of the manuscript from several people; most of all from Sally Thomason and Maarten Kossmann, but also from Peter Bakker, Gerrit Dimmendaal, Frits Kortlandt, Gérard Philippson, Norval Smith, and Caroline Smits. I thank Matthias Brenzinger and Frans Huizenveld for making crucial sources available to me. Many more people were influ­ ential through their feedback after my talks on Ma'a in Bayreuth, Hamburg, München, Leipzig, Leiden, Amsterdam, and Manchester. I am grateful for the support from the University of Dar es Salaam and in particular from the late Ruth Besha who made her MA thesis on Ma'a available for me, and also to the Commission on Science and Technology for the research clearance that they granted me. The most pleasant periods were doubtless the hours that I spent with all the Mbugu, Shambaa and Pare, who shared their insight into their languages and their views on their history with me. In the first place I would like to thank my host Ernest Mbondei and my main informants Dominicki Ndala, William Mboko, and Raphael Mboko; and also William Alois, Yambazi Immanuel, Mzee Ishika, Tito Musa Ishiki, Shabani A. Kalutu, Mzee Abisai Madeghe, Amina Mavare; Salimu Mavare, Bi Mizighi, Bi Mkengwa, Joseph Mcharó Pius, Bw. & Bi. Sabuni; Saidi Salimu, Saidi Mdoki Seamashi, Joseph Singano; as well as many elders in various places in the Pare mountains and in Handeni.

XVII PAS PAST PERF PF

Pl POS

poss pro PST Q

ques Qw Ra RC REFL REL

S Sa sg Sh Som SPEED ST

passive past perferct perfect plural possessive possessive pronoun past quenstion intonation question word Qwadza Rangwi reciprocal reflexive relative subject Sag(h)ala singular Shamba(l)a Somali speed (immediate action) stative

STILL SUBJ

subj Sw T&B74 V

v ap V CS

v ps v rc v sp V St

v:ap WR Zi

still (unfinished action) subjunctive subject Swahili Tucker and Bryan (1974) verb verb with applied extension verb causative extension verb with passive extension verb with reciprocal extension verb with separative extension verb with stative extension verb with lexicalised applied extension, etc West Rift (Southern Cushitic) Zigua = Zigula

XVIII

Map I: Map of the relevant area of East Africa (Names of peoples are in capitals. Map after Brenzinger 1987.)

XIX

Map 2: Map of the Usambara area

1

/. /

INTRODUCTION

Th e people an d th eir language The Mbugu or Ma'á people live in the Usambara mountains in Tanzania. As a people, they speak two languages. However, one of the main conclusions of this book is that these two languages share one grammar and that the vocabulary is parallel. Thus, we are dealing with two languages that are intrinsically linked. The "normal" Mbugu language is very similar to the Bantu language Pare (also called Chasu) both in grammar and in lexicon. There are differences between "normal" Mbugu and Pare (Chasu) but whether these go beyond dialectal variation is difficult to ascertain. Pare speakers are able to understand Normal Mbugu, be it with some difficulty, but charac­ terise it as deviant. I will present "normal" Mbugu and Pare (Chasu) as different but closely related languages. The "inner" Mbugu language (or Ma'a) has a lot of deviant lexical material which is partly Southern Cushitic in origin, while it does not differ in grammar from "normal" Mbugu; it is a parasite of Normal Mbugu. It is this variety that has made it to the linguistic literature as a typical example of a mixed language. The double nature of the lexicon is also exemplified by the two autonyms for the people, vaMbugu in Normal Mbugu and vaMa'á in Inner Mbugu. Outsiders call the people waMbugu. The same roots with a different noun prefix are used by the speakers for the language (in both varieties), thus kiMbugu and kiMa'á. The varieties are distinguished by the speakers themselves only when discussing the differences. In that case they speak of "normal Mbugu/Ma'á" and "inner Mbugu/Ma'á" and I will use these designations, Normal Mbugu (NMb) and Inner Mbugu (1Mb), for the two varie­ ties, alongside the usage of the established name Ma'a for Inner Mbugu. Following Tucker and Bryan (1974:191), Thomason and others have used the name Ma'á for the mixed variety and Mbugu for the plain Bantu variety. This usage has had the unfortunate effect that the intrinsic relationship between the two was played down. In fact, the above presentation of the state of affairs reflects the conclu­ sions of my study and not the common opinion prevalent when I started my research. Tucker and Bryan talk of two subdivisions of a people of a single ethnic unit: the Mbugu speaking a Bantu language Mbugu, similar to Pare, and the Ma'á speaking "their original language, which consists of a non-Bantu substratum upon which a Bantu superstructure has been imposed" (Tucker and Bryan 1974:190). Both the syn­ chronic and diachronic aspects of this view do not hold. By describing both languages in detail I show that they share one grammar and from this corrected view on the syn­ chronic situation and a fresh extensive etymological study, I re-examine the historical conclusions that can be drawn. There is a logical possibility that what I view as a point that earlier researchers missed, i.e. the fact that there are two varieties and that these are interrelated, is in fact a relatively recent phenomenon and that the sociolinguistic situation earlier in the 20th century was markedly different from what it is now. I ar­ gue against this in section 3.2.

?

CHAPTER 1

Culturally and physically the Mbugu people are clearly different from the other Bantu people in the Usambara mountains, i.e. the Shambaa, who are dominant in the area, and the Pare immigrants. Despite long term acculturation, cattle keeping is still associated with the Mbugu and banana culture with the Shambaa. The Mbugu also show differences in physiognomy and in hair structure. It is because of these differ­ ences that the tacit assumption is that once the Mbugu, or at least some of their forefa­ thers, spoke a language that was not Bantu. Whether parts of such an earlier nonBantu language can be traced in the present-day language is a matter of debate which is addressed in 3.5. Like the Pare, the Mbugu are immigrants in the Usambara mountains. They too come from the Pare mountains where there are still place names such as Mbughu-eni (Mbugu-LOC) and where there are still people to be found who know that they are of Mbugu origin although in language and culture they are now indistinguishable from the dominant Pare. The resemblance of the Mbugu language to Pare, both in grammar and in lexicon, is a reflection of that earlier period of residence in the Pare area; pres­ ently the dominating surrounding language and the source of new borrowings is Shambaa. The name Mbugu is not to be confused with Mbugwe.1 Quite far from the Usambara and the Pare mountains, near Lake Manyara, at the other edge of the Maasai plains there is a small ethnic group whose language is called Mbugwe, or Bowe as the speakers say themselves. This language is closely related to Rangi (=Langi) and be­ longs to a different group of Bantu languages, cf. Nurse and Philippson (1975). Whether the similarity in name is a coincidence or not, is a question that will be ad­ dressed in 2.10. 1.2

Earlier descriptions and history of Ma 'á studies In this section I present a short history of Mbugu studies; Brenzinger (1987) contains a more detailed critical overview of studies on Mbugu. I also briefly discuss Shambaa and Pare studies. Mbugu studies started relatively early. The first publication containing Mbugu linguistic data dates from 1885 when Bishop Farler published a short word list. The list, which is phonetically difficult to interpret (see section 3.3), contains words from both Inner and Normal Mbugu without distinguishing them. Its historical value lies in the fact that it contains some archaic words such as the word for 'elephant'. Before him, others, such as Krapf (1858) and New (1873), had already mentioned the Mbugu as a deviant ethnic group among the Shambaa without giving linguistic material. Farler is very explicit in his observations on racial and cultural dif­ ferences between the Mbugu and the Shambaa when he describes the Mbugu as fol­ lows: "They were spare in frame, features rather Semitic than Negroid, being thin and long in the face with thin lips, straight noses, not in the slightest degree patulous; for blacks, very good-looking, intelligent fellows. They have no traditions; they wear only skins; they do not cultivate, but live by breeding cattle". These were followed by simi­ lar observations by Johnston ( 1886), Baumann ( 1890, 1891) and Meyer ( 1890). It was probably these reports that inspired the experienced Bantu-linguist, Carl Meinhof, to visit Usambara on his journey to East Africa; he published his short de­ scription of Mbugu (mostly Inner Mbugu) in 1906. He was the first to pay attention to the mixed nature of the lexicon and he pointed out its importance for historical linguis­ tics, stressing that the Mbugu language presented a rare case in which the lexical roots

INTRODUCTION

3

are apparently non-Bantu while they still take Bantu prefixes in many cases and claim­ ing that this Bantuisation process shows how a language that belonged to one family can be restructured by encapsulation in a different grammar.2 The prevalent opinion at that time was that the Mbugu were some sort of Maasai (Kwavi) and Meinhof gave the linguistic evidence for this Maasai link, but in addition he showed a Southern Cushitic connection in the cognates with Burunge, a language that he had studied on the same journey. According to him the oldest lexical layer was a West African lan­ guage like Ewe. This 'wild' proposal was not followed up by any of his successors. Judging from the forms of certain words Meinhof probably worked on the Bumbuli dialect of (Inner) Mbugu; in fact he does mention the mission station in Bumbuli. His material is mainly from the Inner Mbugu lexicon, but sometimes from the Normal Mbugu lexicon and for some words he has both forms, e.g., 'rain' mare (1Mb) and mbua (NMb). He did not, however, make mention of two varieties or two languages. Before Meinhof, Lieutenant Storch had published a Mbugu word list containing only Inner Mbugu forms, but this publication, Storch (1895), in a way, did not play a role in the history of Mbugu studies because it was ignored by Meinhof and his suc­ cessors. Storch was apparently unaware of Fader's wordlist since he makes no refer­ ence to it and this is not surprising since nothing in the title of the book that contains these data would alert the reader who is interested in Mbugu.3 It is more surprising that the next author on the scene, Meinhof, and all subsequent authors, did not use Storch while making use of Farler. As a consequence we have two independent lexical sources of this early period but, unfortunately, for both the phonetic interpretation is not straightforward, see 3.3. Copland (1933-34) published a short text in Inner Mbugu (only one Normal Mbugu word), which is, until the present work, the only published text available. The text does not deviate from the Magamba dialect of present-day Inner Mbugu. Greenberg (1966)[=1955] classified Ma'á (Inner Mbugu) as a Southern Cushitic language without presentation of the evidence. I assume he considered Meinhof's Burunge cognates to be crucial, his Maasai cognates to be later borrowings and his WestAfrican cognates to be speculation. Since Greenberg, the dominant opinion is that the original Mbugu language was Southern Cushitic. Green (1958) gives some interesting ethnographic information and a very short sketch of the grammar. He also noted down a wordlist which was never published but was used by his successors, including myself. His view on the nature and history of the Inner Mbugu relies heavily on Meinhof. Tucker and Bryan (1974) is based on fieldwork by Bryan in 1958 and by Tucker in 1965 in different places, mainly Magamba but partly with schoolchildren outside the area. This 20-page article is the most detailed account of the language thus far. Tucker and Bryan are the first to describe both Inner and Normal Mbugu and to distinguish these two, without, however, realising their intrinsic connection. Whiteley (1960) reports that Ma'á is a counter-example to claims made by Hans Wolff and Joseph Greenberg that mixed languages in which entire subsystems come from completely different languages do not exist. Whiteley had done extensive fieldwork on Iraqw and apparently he had done also some fieldwork on Ma'á although he takes his Ma'á data from Green and Tucker and Bryan. He suggests a gradual adapta­ tion from a Southern Cushitic language to a Bantu language.

4

CHAPTER 1

Heine did fieldwork in the Mbugu area but he did not publish his data. Instead he discussed the problems that mixed languages such as Mbugu pose to language clas­ sification, related to the problem of primacy of morphology or lexicon in genetic rela­ tionships. Since Wolff (1965) had argued for more rigor in the use of terms such as genetic relationship, sub- and superstratum and "Mischsprache", Heine (1969) pointed out that Mbugu is a real mixed language. Weimers (1966) discusses Ma'á in his influential introductory book to African linguistics. Goodman (1971) makes Ma'á known to the wider linguistic audience and specifically draws the attention of creolists to the "strange case of Mbugu". In his sce­ nario there were two languages; one a Southern Cushitic language that adopted more and more Bantu words and secondly a Bantu language that adopted more and more words from a Southern Cushitic language. The two became more and more alike until they were little more than stylistic variants: Bantu and non-Bantu alternatives existed for a number of words, but gradually the non-Bantu forms were favoured, perhaps to make the language less comprehensible to the surrounding Bantu peoples. I will de­ fend several elements of his analysis although my view on the interdependence of the stylistic variants is slightly different. Goodman argues that if Mbugu were merely a relexified Bantu language then the division of Bantu/non-Bantu vocabulary would be different and one would not expect, for example, that the lower numerals are nonBantu (i.e. relexified) and the higher numerals still Bantu. In 3.12, I will argue that the construction of a parallel lexicon for identity and secrecy is indeed to be expected pre­ cisely there in the core vocabulary. Besha, born in the area, also did fieldwork on Ma'á for her unpublished MA thesis. She is the first to pay attention to the fact that Inner Mbugu and Normal Mbugu words are interchangeable: "There are those Kimbughu [Normal Mbugu] words which are used side by side with Kimaa [Ma'á] words" (Besha 1974: 169). Ehret did extensive research on Ma'á but his data are buried in the tables of his 1980 lexical and phonological reconstruction of Southern Cushitic. How a different synchronic analysis of the Ma'á language leads to a different approach of the recon­ struction of Southern Cushitic is discussed in Mous (1996). Some of Ehret's etymolo­ gies are discussed in the etymological word list (Appendix) and his non-Bantu gram­ matical elements in section 3.5. Thomason (1983) is a reaction against Greenberg's classification and a plea not to classify languages such as Ma'á for which the chain of transmission is broken. She emphasises the non-Bantu grammatical elements taken from Ehret and accepts both Greenberg's and Ehret's claim that the lexicon is of Cushitic origin. In perhaps what can be regarded as a comparable case, Boretzky (1985) has proposed that the mixed language Anglo-Romani developed when gypsies who lost most of their original language (Romani) at the expense of English consciously at­ tempted to restore the original language, i.e. shift plus unsuccessful shift back. Brenzinger (1987) adopts this idea for Ma'á, to which Sasse (1992) adds that "hybridi­ sation" such as in Ma'á is a relexification of the target language after the shift had al­ ready taken place and when vocabulary of the abandoned language was still present. This vocabulary was inserted into the morphosyntactic frame of the target language. This work shows in greater detail how such a process developed. Thomason, in Thomason and Kaufman (1988), sees Ma'á as a Cushitic lan­ guage which is heavily bantuised (stage 6 on the borrowing scale), since the grammar

INTRODUCTION

5

has been replaced by Bantu. "Enough is known about the history of the Ma'á speakers to establish firmly our claim that the mixture in this case resulted from borrowing in a situation of language maintenance, rather than from shift." Whether the knowledge about the history of the Ma'á speakers could possibly allow other interpretations is discussed in section 2.1. There are no ethnographic or historical studies dedicated solely to the Mbugu. Mbugu is simply mentioned in those on the Shambaa and the Pare. Thus an ethnohistorical study on Mbugu would be a most welcome addition to this linguistichistorical study and may prove indispensible in answering some of the research ques­ tions raised in this work. The surrounding Bantu language Shambaa which is spoken all over the Usambara mountains but not outside this area, is relatively well described. There is an ex­ cellent grammar by the German missionary Röhl (1911) and an extensive dictionary by LangHeinrich (1921), which is not tone marked despite the fact that Röhl had dealt with tone in his grammar. My linguistic comparisons with Shambaa will be based primarily on these two sources; occasionally I will make use of more recent studies such as those by Besha (1989), Besha and Nurse (1979), and Odden (1982). On the historical and ethnographic side too, the Shambaa have fared better, with attention from the excellent historian Feierman (1974) and Winands (1962). There is some dia­ lectal variation in Shambaa, for example, some areas show intervocalic deletion of /; hence the language and the people are referred to as Shambaa or Shambala. In the co­ lonial period, Shambaa was used in schools, and material to learn to read and write Shambaa was available (see Besha 1989:1-7). Pare, known by the speakers themselves as Chasu, is spoken on three mountain ranges north of the Usambaras, commonly referred to as South Pare, Mid Pare and North Pare. The dialectal variation within Pare is considerable. As stated before, Pare is very important for the study of Mbugu. Unfortunately, the state of linguistic de­ scription of Pare is far from satisfactory. The most important sources are the short de­ scription by the German missionary Kotz published in 1909, the articles by KählerMeyer on tone in 1962/63 and 1964, the book by Kagaya (1989), and the Ph.D. thesis by Abel Mreta (1998). The grammar by Kotz is not tone-marked although he did rec­ ognise the distinctive value of tone (p. 3) and it does not contain a vocabulary. KählerMever's articles are excellent but deal only with some aspects of grammar. Odden (1986) has used these articles as the basis for a phonological reanalysis. Kagaya (1989) contains vocabulary in the form of the standard Tokyo questionaire and a grammatical sketch. He used speakers from different dialects and noticed tonal differ­ ences between these dialects. My primary source for lexical data is an unpublished collection by Philippson (n.d.) because the tones from his dialect are closer to the ones that I noticed for Mbugu. Mreta (1998) is invaluable for an understanding of the tense/aspect system. The history of the Pare has been dealt with in great detail by the Tanzanian historian Kimambo (1969). 1.3

Source of data, fieldwork This study is based on a series of three fieldwork periods of a few months each in the Usambara mountains in the years 1992-1993. The fieldwork was conducted in Swahili. I started in the Magamba area working on Ma'á (Inner Mbugu). First, I elic­ ited short sentences and collected a few historical texts in Ma'á from several older

6

CHAPTER 1

people in the Magamba area. On several occasions I was promised folktales in Ma'á but in the end I only got one such story from my main informant William Mboko. Later I travelled to the area to the east as people claimed that the Mbugu of that area had a preference for Normal Mbugu. It was then, when I used the same short sen­ tences for elicitation, that the true relationship between the two varieties occured to me. There too I collected some texts in Normal Mbugu, some of which turned out to be a mixture of Normal and Inner Mbugu. These texts are historical and cultural in nature and were based on interviews. I worked on the lexicon with a variety of speak­ ers using the words that appeared in elicitation and in texts as a basis. During later fieldwork periods I visited the other Mbugu areas such as the Rangwi and the Bumbuli area. The nature of my fieldwork was by no means a dialect survey. It was only in the Magamba area that I worked with a variety of speakers and most variation is found in that area. The Magamba area is much bigger than Bumbuli and Rangwi. My fieldwork notebooks contain roughly 315 pages with material from elicitation, 200 pages of tran­ scribed texts (transcribed with Mbugu speaking assistants) and 215 pages of lexical notes. I worked in Magamba with Dominicki Ndala, William Mboko, Raphael Mboko, Salimu Mavare, and specifically on texts with Joseph Mcharó Pius; Mr. and Mrs. Sabuni; Mrs. Mizighi, Mrs. Mkengwa, Mr. Ishika, and on the lexicon with Amina Mavare; in Malibwi with Shabani A. Kalutu, and on texts with Mzee Ishika and Mzee Salimu; in Rangwi with Mzee Abisai Madeghe, Yambazi Immanuel, Saidi Salimu and specifically on texts with William Alois, Joseph Singano; and in Bumbuli with Tito Musa Ishiki, and Saidi Mdoki Seamashi. 1.4

Areas of settlement of the Ma 'á/Mbugu and clan affiliation All Mbugu meet Shambaa people in their immediate environment and most Mbugu have Pare neighbours as well. The Mbugu mainly live in three pockets in the Usambara mountains. One encompasses the villages Rangwi, Fwizay, Emao, at the north-western edge of the mountains, which I will refer to collectively as the Rangwi area; a very small pocket consists of the villages Tetei, Magila, Lughangé and Kwehangala (=Mpalái), all close to each other and situated between Soni and Bumbuli. I will call this area Bumbuli. The largest Mbugu population is to be found in an area roughly between Magamba and Malibwi, north of Lushoto, which I call the Magamba area. These areas can be seen in map 2. It is important to add that there is no area in the Usambaras which is exclusively inhabited by Mbugu. Despite the relatively small numbers of Mbugu speakers in an area where dis­ tances are never more than a day's walk, the dialectal variation is substantial. This variation is phonological and lexical. The phonological variation is basically among the velar consonants x, k and h, see section 4.2. The lexical variation is in Inner Mbugu; the Bumbuli area shows more lexical differentiation when compared to the others. Here, for example, there is a distinction 'cow' versus 'cattle', de versus wa, reminiscent of Iraqw (Southern Cushitic) hlee versus hikwa, and a distinction between human and animal blood. Apart from the dialectal variation, there is also variation in proficiency levels and in language preference for Inner Mbugu. In the Rangwi area there is a concentra­ tion of people who do not speak Inner Mbugu or Ma'á; in the Magamba area on the Malibwi side there is a preference for Normal Mbugu. These are also the areas where the concentration of Pare speakers is higher when compared to the other areas, but the

7

INTRODUCTION

main conditioning factor behind this variation is clan affiliation. Thus, there is a corre­ lation between clan affiliation, language preference and settlement. This correlation is not to be seen as absolute; obviously there is individual variation. The tendencies, however, are of potential interest for the reconstruction of the history and will be dis­ cussed in more detail in section 2.1. The following table summarises the correspon­ dences in the distribution of clan, area, and language preference. The MagambaMalibwi area is one area only differentiated here to indicate a gradual difference in clan and language preference. The order of 1Mb (Inner Mbugu) and NMb (Normal Mbugu) in the third column reflects my impression of language preference for the relevant area. Area Magamba Magamba-Malibwi Rangwi Rangwi Bumbuli Bumbuli

Clan Londo Nhkandu Ombeji Gonja Nhkangwana Lomboza

Language 1Mb, NMb NMb, 1Mb 1Mb, NMb NMb 1Mb, NMb NMb, 1Mb

Table 1.1: Area, clans, and language preference 1.5

Multilingualism and language attitude Apart from Mbugu, the children tend to learn three other languages before they go to school: Shambaa, Swahili and Pare. This order reflects the average degree of proficiency. Pare should be the easiest language for a Mbugu person to learn but it is also the least important one. The acquisition of Shambaa is inevitable in the area. Shambaa people reported to me that they can easily spot a Mbugu when he speaks Shambaa; they speak too slow, make no elisions and make mistakes in the tones. Communication between Mbugu and Shambaa is usually in Shambaa inspite of the neutral option of speaking Swahili. Shambaa people do not speak Mbugu. Speakers of Inner Mbugu invariably also speak Normal Mbugu. A sociolinguistic survey among the Mbugu was carried out by the Summer Institute of Linguistics. They worked on the assumption that Inner Mbugu (Ma'á) and Normal Mbugu (Mbugu) are two different languages and one of their findings was that those who speak Inner Mbugu also speak Normal Mbugu but not always the other way around. In principle, it would be possible to find speakers who know only Inner Mbugu and not Normal Mbugu but these are either extremely rare or do not exist at all. One can speak Inner Mbugu by degree, i.e. the amount of parallel lexicon that one uses or that one is able to use varies from individual to individual. People who answered in the question­ naire that they do not speak Inner Mbugu, know too little (or nothing) of the extra lexicon, or are not confident enough about it to claim proficiency. The other way around, people who claim to know Inner Mbugu and not Normal Mbugu, seems to be rare. They will always know not only the Normal Mbugu grammar but also a consid­ erable part of the vocabulary that is shared and felt to be Normal Mbugu. To know Inner Mbugu imperfectly or to know Normal Mbugu imperfectly is not symmetrical to the speakers because speakers would feel they know Inner Mbugu imperfectly if they do not know some elements of the Inner Mbugu lexicon; whereas the other way

8

CHAPTER 1

around, if someone knows only an Inner Mbugu lexeme and not the Normal Mbugu parallel, this is not felt as knowing Normal Mbugu imperfectly. In Tetei in the Bumbuli area, where there are no Pare neighbours and where all Mbugu are confident Inner Mbugu users, children under ten years of age, just like their parents, had no trouble telling me which word is Inner Mbugu and which is Normal Mbugu and they could always give the Normal Mbugu counterpart of an Inner Mbugu word. Thus, even there, both varieties are acquired in early childhood. The use of the Inner Mbugu (Ma'á) lexicon is by no means restricted and it is important to note that in many families Inner Mbugu (Ma'á) is the norm and switching to Normal Mbugu is a marked choice. In earlier works, I have used the term "register" for the parallel lexicon called Inner Mbugu or Ma'á. In Mous (2001a), I have called it an ethnoregister and although this captures the structural dependency of Inner Mbugu on Normal Mbugu and the parallelism in the lexicon, the term has its problems and I am now convinced that the use of the term is misleading for several reasons. First of all it should be mentioned that there are many speech situations, families, wards or villages in which Inner Mbugu or Ma'á is spoken as the unmarked choice, and some people only seldom use the basic "register" (Normal Mbugu). Obviously Inner Mbugu can and is used for all topics of conversation and it is as complete in its lexical range as Normal Mbugu. The term "register" refers to language use as a situationally defined variety, that is, as a marked choice and thus, a register in the strict sense has a restricted functional range (Ferguson 1994: 20, see also Biber 1994:51-53, for a discussion of the term register). Thus, given this definition it is not proper to call Inner nor Normal Mbugu "registers". Mbugu people's attitude to and future expectations for their language (Inner or Normal) Mbugu are positive and confident. There is no breakdown in transmission and the children learn their parents' language. The language is not threatened by extinction in the near future. There is however a decline in the extent and knowledge of the Inner Mbugu lexicon among younger speakers, see section 2.2. The few grammaticalisation processes that can be observed derive from Normal Mbugu lexical forms and not from Inner Mbugu lexicon, see section 5.2. 1.6

A sample of Mbugu language material In this section I present a short piece of a text in order to give a first impression of the Mbugu language and the differences and interrelatedness of Inner and Normal Mbugu. This short excerpt consists of the initial sentences from a short "historical" text that was narrated to me in the Inner Mbugu variety by William Mboko. All lexemes that belong to the Inner Mbugu parallel lexicon are marked in italics and their Normal Mbugu counterparts are indicated immediately above them. Replacing the lexemes in italic by those above them results in a Normal Mbugu text. Only for two lexemes the narrator used the Normal Mbugu form, hand 'plant' and twai 'there', where he could have used an Inner Mbugu form. This is typical for a story told in Inner Mbugu, that is, when one speaks Inner Mbugu, one does so maximally but seldom completely.

INTRODUCTION

9

Sample text 1 hé -na m -zima é -tang -we kimwérí m- fumwa w-a -we kimwérí dilaó w-a hé -ló m w -agirú é -sé 16-havel -elder 1 -call -PAS:PF Kimweri king 1-CON 2 i-i i -sanga 1-á lusótó. yá i -di 1-á lusótó. this 5-land 5-CON Lushoto. 3 hé -na hé -ló 16-have

i -zuva i -'azé 5 -day

i -mwe i -wé 5 -one

áa -tanga áa -sé l:PST-call

4 va-mbugu na va -shamba na va-Ma'á na va -sita na 2 -Mbugu with 2 -Shambaa with 5

va -asu va -'ariyé 2 -Pare

vá -vata vá -zé-m -h/ -ía ma -diyo -ake vá so vá -zé-m -hand -ía ma -gerú kú'u. 2:SBJ -go 2:SBJ-IT-1 -plant-APL 6 -bananas his

6 íji va -mbugu v-á- he -fika ila'í íji va -Ma'á v-á- he -hé twáí. now 2 -Mbugu 2-PST-16-arrive there 7 kwá kubá te-vé- kund -ye vá -ronga io ndima kwá kubá te-vé- dúmú -ye vá -bó'i ká nyamalo, with reason NEG-2- want -PF 2:SBJ -make DEM2 work.9 "There was an elder called Kimweri, king of this land of Lushoto. On a certain day he called the Mbugu, Shambaa, and Pare people to go and plant his banana trees. Well, the Mbugu arrived there. Because they didn't want to do this work, ..." The continuation of the story recounts that the Mbugu planted the banana trees upsidedown. Upon which the chief concluded that the Mbugu were stupid and useless and refrained from calling them for communal work from then on. The story illustrates both the fact that the surrounding people look down upon the Mbugu and that the Mbugu claim not to be bothered by this. The lateral fricative hl in vázémhlia 'that they planted' in the fifth line of the sample text, and the glottal stop ' in va'ariyé 'Pare' in the fourth line are sounds that are limited to the Inner Mbugu lexicon. The sample text, being the beginning of a story, does not show the richness in the tense system since the only "tenses" are the perfect (PF), the past (PST), the itive (IT), and the subjunctive (SBJ); the text also illus­ trates verbal derivation such as the passive (PAS) in the first line and the applied (APL) in the fourth line. The text already shows that parallel nouns are grosso modo in the same noun class, with the same noun class prefix, dilao or mfumwa 'king'in the first line being an exception. Function words are parallel too, see the demonstrative DEM2

10

CHAPTER 1

in the last line, and 'his' in line five. Agreement is slightly different in Inner Mbugu: the demonstrative in the last line does not show agreement with the (following) head noun in Inner Mbugu but it does in Normal Mbugu; similarly, the possessive in line 5 shows agreement with the (preceding) head noun in Normal Mbugu, but for this par­ ticular noun class (class 6) the agreement morpheme is zero. Other function words are not parallel, such as the genitive pronoun glossed with CON in the first two lines, the multifunctional na glossed with 'with' in the third line and kwa kuba in the last line. Note that the names in this text do not permit a choice between Inner and Nor­ mal Mbugu, a property which extends to all names. The text illustrates that the Mbugu language consists of one structural system with two interrelated lexical systems and that the impact of the parallel lexicon, Inner Mbugu, is so far-reaching that one can tell a story without using lexical roots from the Normal Mbugu vocabulary. 1.7

Properties ofthe parallel lexicon The grammatical differences between Inner Mbugu and Normal Mbugu are minimal. The two differ in terms of their the lexicon but these lexical differences in­ volve core vocabulary and even closed sets of function words such as personal pro­ nouns, possessives, demonstratives, etc. Ma'á or Inner Mbugu is in fact a parallel lexi­ con to the Bantu language (Normal) Mbugu. A parallel lexicon is the structure of (part of) a lexicon in which two word forms are on a par, that is, the two word forms share meaning, metaphorical extensions, and morphological properties such as noun class membership for nouns and predicate frame and derivational suffixes for verbs. The parallel sets are also coherent in the sense that speakers know to which of the sets a particular form belongs and consequently also when to use which form. The lexicon of Ma'á is parallel to that of Mbugu; in the sense that Ma'á forms are on a par with Mbugu forms sharing semantics and formal morphological properties. In (1.1) we give a few examples of not so common lexemes that have the same meaning in both lan­ guages. Both languages make the same semantic divisions of lexical domains, see (1.2) for the domain of 'getting up, leaving'. Some examples that the two languages show the same metaphorical extensions for the same lexemes are given in (1.3). (i-

Ma a meaning fili to go out of sight over the hill sibi strain (beer) 'ihlé unripe, bad

(1.2) Ma'á ku-ká ku-dáha ku-só ku-dóda

meaning to get up to walk, to step to leave, to go to to go for a walk

VÍSÍ

meaning to go out of sight over the hill strain (beer) unripe, bad

Mbugu ku-vúka ku-vata ku-tónga ku-anga

meaning to get up to walk, to step to leave, to go to to go for a walk

Mbugu gongolóka chumbua

11

INTRODUCTION (1.3) Ma'á mxatú hlúku hi

meaning tree, afterbirth to drop, give birth to sew, mould

Mbugu mtí gwisha chuma

meaning tree, afterbirth to drop, give birth to sew, mould

Parallel nouns by-and-large have the same noun class membership, (1.4); noun class membership is a formal morphological property that is not determined by the seman­ tics of the noun. (1.4) Ma'á mhé vahé mharégha miharégha vubasá mabasá lunábo nábo lumuhlú muhlú lupaé mhpaé vuhuyé

meaning person class 1 people class 2 arm class 3 arms class 4 face class 14.1 faces class 6 rib claw 11 ribs class 10 rope class 11 ropes class 10 saliva, curse class 11 saliva, curses class 10 sorrow class 14.1

Mbugu mnhtu vatu mkóno mikóno vusho mashó luvaju mbaju luzighi ndighi lute nyóté vukiva

meaning person class 1 people class 2 arm class 3 arms class 4 face class 14.1 faces closs 6 rib, lass 11 ribs class 10 rope class 11 ropes class 10 saliva, curse class 11 saliva, curses class 10 sorrow class 14.1

Verbal derivations are parallel in both languages and this is also true for lexicalised cases which contain frozen instances of causatives, =ti, =ij, and passives, =w ((1.5)). Parallel verbs have the same lexical aspect (Aktionsart) as is shown in (1.6) where the expression of the state of tiredness needs a past tense in both languages. Parallel verbs have the same argument structure, for example, the body part object of 'to break' in (1.7) and (1.8) is used without a possessive marker in both languages. (1.5) Ma'á ge=wa

to be opened, to be allowed to get married yó'=íja accuse 'ú=ti cure sanúí=wa be angry makáhlú=wa dream

Mbugu jughul=wato be opened, to be allowed to get married tét=íja accuse k-ija cure shin=wa be angry ósh=wa dream

(1.6) Ma'á n-åa-kwaha 1 SG-PST-tired 'I am tired'

Mbugu n-áa-ghata 1 SG-PST-tired 'I am tired'

(1.7) åa (-kú) -på kusåme 1:PST -RFL -break 15.leg. 'He broke his leg.'

(1Mb)

12

CHAPTER 1

(1.8) áa (-kú) -haha 1:PST -RFL -break 'He broke his leg.'

kughú 15.leg

(N M b)

One of the important linguistic properties of lexical manipulation is that the re­ placing word has the same morpho-syntactic properties as the word it is replacing. Thus there is a strong tendency for parallel borrowings to have the same noun class, as Kunene observed for the hlonipha register of respect for in-laws by name avoidance, "It would seem, however, that an attempt is always made to employ a substitute noun (or coin one) in the same class as the one which is being replaced" (Kunene 1958: 167). In the following examples of Southern Sotho hlonipha supplied by Kunene, the replacing words from Afrikaans receive the appropriate noun class prefix, a class 5 CV noun class prefix le in these cases: lehlanya 'madman' is replaced by lekopo from Afrikaans malkop4 lekhooa 'white man' by lebas e from baas 'boss', litelu 'beard' by li-barete from baard, letsha 'dam' by letamo from dam (Kunene 1958). In the Nyakyusa register of respect ili-paamba (5/6) 'brick' is replaced by i-tofali from Swahili tofali and i-sekema (9/10) 'fever' is replaced by i-homa from Swahili homa, both with addition of a noun prefix i- to adapt to the Nyakyusa noun class (Kolbusa 1995). Simi­ larly, in the argot of Yibir magicians among the Somali: iftin 'moon' is masculine just like Somali dayax 'moon', while iftin 'sun, rupee' is feminine just like Somali qorrax 'sun', both derived from ifiin (m) 'light, brilliancy'. The Arabic borrowing albaqar (f) 'cows' has the gender of the Somali equivalent(s) lo ' or sa (f) and the loan katowa (m) 'mouth' has the gender of Somali af (m) 'mouth' (Banti 1997). This property of shar­ ing the same noun class is a strong indication of lexical manipulation and this is a property of the Ma'á lexicon as evidenced in (1.4) above. The study of Ma'á can be placed in the context of the wider phenomenon of parallel lexicon formation by looking at the processes that may lead to such a parallel lexicon. This is what I have done to some extent in Mous (2001a and b, to appear). Conscious parallel lexicon building or lexical manipulation occurs in many situations. Situations in which they occur are for example: registers of respect, taboo, argot of social groups, slang, urban youth languages, initiation languages, secret codes, and ritual languages. In all these cases we find the same morphosyntactic properties shared by the parallel lexis. The strategies that have been used in the formation of Inner Mbugu are common strategies for conscious lexicon creation. These include borrowing from a wide range of languages (and into the appropriate noun class) (cf. Table 2.4), the use of archaic words as Inner Mbugu words (cf. section 2.2), truncation of a word while adding a dummy suffix vowel e, substitution by remarkable sounds, and metathesis. Other strategies that are used in conscious lexicon formation in situations of taboo include circumlocution and semantic extension. These are less apt for situations in which the purpose of the manipulation is not to avoid a word but to make a language that sounds different for identity purposes, as is the case for Ma'á. A number of Inner Mbugu forms seem to be derived from the Normal Mbugu form by changing the final vowel to e and changing the tone pattern to L(L~L)H; cf. Table 1.2. Hypothetically, the 1Mb forms could be from another but related source language, but no such language could be identified. If we accept that one is derived

13

INTRODUCTION

from the other, then the direction must be from Normal Mbugu to Inner Mbugu since the uniformity in form is in 1Mb. 1Mb nhkungé makale ndaté mzimè mhiné mnyawé nhkimé kigingé ngilé

NMb nhkungú makáá n dat a mzima mhini mnyáwu nhkima kigíngí ngilá

meaning ankle charcoal stick old man handle cat monkey stump bee

source5 Sh: nhkungu Pa: makáa Pa: mzátá Pa: mzima Pa: mhini Pa: inywáwi Pa: nhkima Pa: kigíngí Origin unknown

Table 1.2: Derivation of Inner Mbugu by dummy suffix Substitution of a consonant by a sound that is deviant from a Normal Mbugu point of view as a strategy to derive an Inner Mbugu word is observed in a restricted number of cases. A small number of Inner Mbugu words seem to be derived from their Normal Mbugu counterpart by changing  to glottal stop. Note that a glottal stop is characteris­ tic of Inner Mbugu and such a change renders a word automatically recognisable as Ma'á. These cases are Inner Mbugu muru'u 'sweat' from Normal Mbugu (and Pare) muruke, and Inner Mbugu tí'i 'lift' from Normal Mbugu (and Pare) tika. The same change can be detected in the word là'a 'wait' (1Mb) from Iraqw/Gorwaa lak-iit. The most striking Ma'á sound, however, is the lateral fricative, h1, and in two cases Ma'á words seem to be derived from Normal Mbugu words by changing a consonant to the lateral fricative: nhxwahlé 'partridge' from Normal Mbugu (and Pare) nhkwaré; and mhla'aré 'funnel' from Normal Mbugu mharé. Occasionally there seems to have been metathesis of a round vowel and a: i-kádo 'ten', cf. Oromo kud'aani, and i-'alú 'sheep', cf. Boraana-Oromo (h)oolaa. Consonant metathesis probably occurred in makahlú 'dream', cf. Dahalo milaak'-ani, and i-geru 'ripe banana', cf. Sonjo (Bantu) e-rigo or Kikuyu (Bantu) i-rigo. These lat­ ter examples could be instances of inversion of syllables which are known to occur in situations of lexical manipulation such as argot (see Niceforo 1912:160,162).6 New insights in this work The aim of this work is first of all the presentation of new data. Earlier discus­ sions of the mixed language Ma'á were based only on grammatical sketches such as Meinhof (1906), Green (1958), and Tucker and Bryan (1974). The major lexical source of Ma'á up till now has been Ehret (1980). It has, however, seldom been used as a source for the Ma'á lexicon since one has to unearth the Ma'á material from the reconstructions. The lack of data has been a major problem in the discussion about Ma'á. The earlier literature contains one short text in Ma'á, (Copland 1933-34). Yet it is only from texts that one can get a realistic picture of any deviant language manifes­ tation. The Ma'á texts show for example that the extent of replacement of lexical roots is so extreme that this fact in itself is a major difference between Ma'á and other phe­ nomena of a parallel lexicon such as argot, or even mixed Romani languages such as Anglo-Romani. 1.8

14

CHAPTER 1

The second major issue of this study is the intrinsic relationship between the mixed language and the normal Bantu language. This relationship (common grammar and parallel lexicon) has escaped the attention of linguists because of the common ab­ straction to study a certain variety as a unit, i.e. the mixed language Inner Mbugu was studied separately from the normal Bantu language Mbugu. Subsequently the discus­ sion has, understandably, always concentrated on the mixed variety. Some previous authors, however, had already hinted at a relationship between the two varieties. Tucker and Bryan, for example, mentioned that Ma'á "grammatically behaves like Mbugu", (Tucker and Bryan 1974:191). Their 1974 article was based on fieldwork by the two authors at different places and these circumstances prevented them from rec­ ognising the true nature of the relationship between Ma'á and Mbugu. Besha (1974:169) remarked "There are those Kimbughu words which are used side by side with Kimaa (sic) words. This happens when at one time an informant is asked about a certain word in Kimaa, and he gives a Kimaa word, but in speech uses another word, a Kimbughu word." She sees this as evidence of a trend to replace Ma'á words but she does not elaborate on this parallelism. Although all the other authors did mention the non-mixed variety, they did not pay much attention to the relationship between the two. Moreover, the names for the two varieties as used by linguists did not contribute to an understanding of their interrelationship, as was already mentioned above. The third important contribution this books offers to Ma'á studies is the recog­ nition of a variety of lexical sources of the Inner Mbugu (Ma'á) lexicon. Discussions of the special case of Ma'á were often based on a simplistic view of a "Cushitic lexi­ con" even if the presence of Bantu lexical material was acknowledged. I show in sec­ tion 2.2 that there are several lexical sources. Two observations are of particular im­ portance: One is the considerable contribution of Maasai, a Nilotic language, and the second one is the fact that there is more than one Cushitic source. The presence of Maasai material in Ma'á was already pointed out by Meinhof (1906) but its relevance was underplayed by subsequent authors or it was simply not considered. In addition, the recognition of more than one Cushitic source is vital for an understanding of the history of Ma'á. A number of the roots that I mark as "Old Kenyan Cushitic", the sec­ ond and older Cushitic source, were proposed by Ehret as inherited Cushitic material even where he could find no cognates with Southern Cushitic languages but only with Eastern Cushitic languages (see his table  in Ehret 1980). I interpret the lexical items with Eastern Cushitic correspondences as a second Cushitic source different from the Southern Cushitic source. This is strongly supported by the fact that a number of Southern Cushitic roots in Inner Mbugu share the phonological innovations that char­ acterise the recent subbranch of Iraqw and Gorwaa (see Table 2.2 and Mous 1996). The exact classificatory position of this second Cushitic source is difficult to establish for reasons discussed in section 3.1. The analysis of Inner Mbugu being a parallel lexicon to Normal Mbugu has its consequences for which lexical elements are investi­ gated. A number of roots in Normal Mbugu are ultimately of Cushitic origin. Since the same loans are found in Pare and other Bantu languages of the area, the origin of these roots is not relevant for the history of Ma'á. Our first concern is to trace the extra roots of the parallel Inner Mbugu lexicon and not those of the Normal Mbugu lexicon.7 Due to the parallel lexicon nature of the mixed language we cannot simply assume that Cushitic lexical material cannot be borrowed and must be inherited as Ehret seems to do.

INTRODUCTION

15

A fourth novel insight is found in the correlations between clan settlement, lan­ guage knowledge (Inner and Normal Mbugu or only Normal Mbugu), route of migra­ tion and dialect differences in Ma'á. The relevance of these correlations is discussed in section 2.1. 1.9

Organisation of this book In Chapter 2, I discuss the historical processes in Inner Mbugu, taking evidence from the documented history, from the dialectal variation, and from the changes in form of transferred lexical material comparing the form in the source language with that in present day Mbugu. Using the analysis of the different lexical sources, as well as oral history, I discuss the most likely scenario of the migration history of the Mbugu people. In Chapter 3, I discuss the linguistic history. After stating the issues at stake and presenting the debate about the linguistic origin of Ma'á, I discuss the his­ torical processes. The non-Bantu grammatical elements are examined as to their prob­ able source. I also look at the question of which non-Bantu elements are remnants of the grammar of an earlier language and which are innovations and in doing so I at­ tempt to find some syntactic and conceptual clues to an earlier language. This chapter is crucial in an evaluation of Thomason's ideas about the history of Inner Mbugu since remnants of Cushitic morphology are crucial for her thesis of gradual bantuisation. Chapters 4 to 9 contain a description of both Inner and Normal Mbugu at the same time showing that they have one common structure. All differences between Inner and Normal Mbugu are mentioned. Starting with phonology, the only difference is the re­ striction of certain sounds to words of the Inner Mbugu lexicon. Verbal morphology (Chapter 5) is the bulk of the description mainly because of the high number of dis­ tinctions in the inflectional domain of "tenses" of the verb, which are all shared by both languages. Nominal morphology, including the noun class system, is explained in Chapter 6. Chapter 7 deals with adjectives. Pronouns include demonstratives and possessives. These function words have double forms too, a Normal Mbugu and an Inner Mbugu form. They follow the Bantu pattern of distinctions, but there is a slight difference between Inner and Normal Mbugu in the agreement patterns. Chapter 8 presents the words that do not fall into any of the other word categories. Chapter 9 is a very brief introduction to the syntax, to standard expressions such as greetings, and to the structure of code-switching. It contains some sample texts. The lexicon is presented in the Appendix. The lexicon contains etymological evidence and candidate cognates. The task of establishing the source for Inner Mbugu lexemes has not been completed by far. The Appendix also contains an English index to the lexicon. In the next chapter I go through the history of the Mbugu people as it can be re­ constructed on the basis of an analysis of oral histories in combination with lexical evidence. This journey took me through the various places where the Mbugu are claimed to have been, in order to collect local versions of the oral histories and the necessary lexical comparative material where published sources on the relevant lan­ guages were insufficient.

2

HISTORICAL AND GEOGRAPHICAL BACKGROUND

2.1

Oral History An excellent overview of the various versions of the oral history of the Mbugu can be found in Brenzinger (1987:41-79). He compares the traditions as written down by Meinhof (1906), Rikli (1929), Copland (1933/34), KiTnambo (1969), Kiinzel et al. (1982). After Brenzinger (1987), Conte collected oral histories in Usambara about the Mbugu and he has summarised and analysed these in his Ph.D. thesis (Conte 1994). I can add to this some oral histories that I collected among the Mbugu in Usambara and about the Mbugu among their neighbours. In December 1993, I visited various places in the Pare mountains, as well as Handeni in the Zigua area, and the Mbugwe area searching for some indications of Mbugu history. Comparing the various versions of the oral history of the Mbugu there are a number of points that are mentioned in sev­ eral versions; these are: 1. 2. 3. 4. 5. 6.

Ultimately the Mbugu are from Laikipia (Lukipia) in Kenya (see section 2.8) They moved to South Pare, referred to as Vudee or Vudei, where they lived for some time (see section 2.6) Diffusion from South Pare was in three directions: Zigua area, North Pare, and south to Northwest Usambara (Rangwi) At some point, before or after the Pare period, they were in the place where the Mbugwe now live (see section 2.5) They were working for the Maasai for a period of time (see section 2.4) They entered Usambara from the North and from the West (see section 2.3)

An exegesis of the various oral histories recorded by other researchers and myself is given in the following.1 There are major differences in the depth in which these sources treat the oral history: Copland (1933/34) and Künzel et al. (1982) are based on one account; Conte (1994), Feierman (1974) and Kimambo (1969) are summaries and analyses of a larger number of accounts. I compare all of them on the same six points given above (maps 1 and 2). Copland (1933/34): 1: Lukupia. 2: Through the "second sea",2 Mbugwe, Moshi (North of the Pare moun­ tains), Same (Pare mountains) to Lasa (the mountain between the North­ ern edge of Usambara and the Southern edge of South Pare). 3: Diffusion from Lasa to Zigua (the Dufu who are now assimilated to Zigua), Vudee in South Pare (the Mbugu), and to Shume in Usambara (the Nango who are now assimilated to Shambaa). 4: Stay in Mbugwe prior to arrival in Pare.

18

CHAPTER 2 5:

Maasai threat evoked this diffusion. No mention of a period spent among the Maasai.

Green (1963) 3: Diffusion from Vudee (South Pare) in three directions: Shume (Usambara), North Pare, Nguu hills in Handeni, some in Pangani district (South-East of Usambara). Feierman (1974:74-76): 1: Lukupia. 2: Through Mbugwe. 3: Complex series of movements: some going to Zigua, some to Pare and eventually to Usambara. 5: At a number of points along the way they fought the Maasai (no mention of a longer stay). Künzel et al. (1982:42-44) quoted from Brenzinger (1987:64-66), based on Nyaki Nschanda, "the last Mbugu chief' in Usambara: 1: Laikipia (=Lukipia). 2: Through Taita hills and Taveta in Kenya, Njoro (Moshi), Chawambugu (Same), to Wanduju (Vudee, South Pare). 3: Dispersal into three directions. One to Handeni; one to Shume and Rangwi; one through Mbaru to Bumbuli. 4: Another oral history, that of the village of Mbwei, mentions Mbugu influx from Mbugwe. 6: Several distinct entries into Usambara. Besha(1974: chapter 1): 1: Kelicho (=Kericho) in Kenya near Laikipia. 2: Through Mbulu (Iraqw area) to the Pare mountains. 3: Dispersal. One group crosses the Pangani river to Handeni and another group from the Pare mountains to Shume. Ultimately the two groups converged in the forest between Lushoto and Shume (i.e. more or less in the Magamba area where most Mbugu still live). 4: Stay in Mbugwe prior to the Pare mountains. 5: Fight with the Maasai was in Kelicho. 6: The group in Handeni split: one to Korogwe and from there to Mnyuzi and Mghambe (Mavumi) or to Bungu hills or to Kerenge and Bumbuli; another one to Mombo, Shambalai, Kidundi, Vugha, Soni, Gare, Ngongei, Kivumo, Bagha, Nkongoi, Kwemakame, Malibwi. Conte 1994(46-49): 1: Laikipia plateau chased away by Maa(sai).

HISTORICAL AND GEOGRAPHICAL BACKGROUND 2: 3: 6:

19

Through Rift valley and Northern Tanzania. Some accounts recall a time of residence among the Zigua. Migration to and residence in Pare. Entered Usambara in separate groups.

The following oral histories are based on my own collections. Malibwi, (East of Magamba), told by Mzee Ishika from the Londo clan, on 13.02.1992: 1: Lukupia 2: Arrived in Vudee. 3: Dispersion from Vudee. The Londo went to Morogoro and later passed through Handeni to Usambara and settled in Magamba and other places; the Nhkandu followed the Nango through Shume; 6: Implied several entries into Usambara Magamba, told by Mzee Sewia, on 02.02.1993: 1: Luhiza 2: Arrived in Vudee. 3: Mbugu in pursuit of their cattle stolen by Maasai. 5: Period of work for Maasai. 6: After stealing cattle from Maasai, through Zigua area into Usambara near Bumbuli prior to Kilindi power. Tetei (near Bumbuli), told by some members of the Nhkangwana clan, on 20.10.1993: 1: Lukupia 2: Through Mbugwe to Vudee. The Mbugu and Mburu (=Iraqw) were neighbours. 5: The Nhkangwana came from the Pare mountains through Zigua. 6: The Nhkangwana (and no other Mbugu) entered the Usambara mountains near Bumbuli The following oral traditions are from Mbugu people in the Pare area: Kimambo (1969:61-62): 2: From the Maasai steppe across the Ruvu river. 3: From Vudee (South Pare) first some went to North Pare, then the major­ ity went to the Usambara mountains. Mbughwini, North Pare, told by Mzee Mariko, on 19.12.1993 1: Sumaria (=Somalia).

20

CHAPTER 2 3:

Wambugu came through Gonja (S. Pare), Vudee (S. Pare), Lembeni to Mbughwini (N. Pare). The Mbugu arrived later in North Pare than the Washana (Gweno) and the Wasuyia (Pare).

Kisiwani in South Pare, told by seven elders from the Mbagha clan on 20.12.1993 1: Iraq. 2: They lived in the bush near Mbugwe (war with Maasai). 3: Entered South Pare at Mghende, Gonjanza and Makanya. Those from Mghende dispersed after stealing cattle from Maasai. Some went to Vudee, some to Mbagha, some to Gonja (all South Pare), others went to Sambani (Usambara mountains). 5: No mention of a period in Handeni. Muheza in South pare, told by Mzee Idi Salim Malaki on 20.12.1993: 2: Originally from Mbugwe where they fought with Maasai and others about cattle. 3: They went to different places. People of Mbugu descent can be found among Maasai, Pare, Zigua and others. Senkoro in Changoko-Makanya, told by a meeting of about 20 elders on 24.12.1993: Syomalia (=Somalia), near Tomakisi (unidentified place) They stayed in Mbugwe as cattle-keepers Dispersal from Vudee after war with Maasai; one group went to Chan­ goko-Makanya. 5: They went as far as Losogonoi (Maasai Plains). Summing up: For point 1, the ultimate origin, several places are mentioned. Oral his­ tories usually start with the ultimate origin and this is a political statement about what the people consider to be essential for their identity.3 If the Mbugu mention Somalia, this is a statement that they consider themselves to be different from the other peoples in the area and they mention Somalia as a symbol of Cushitic-ness or "Hamitic"-ness, a label that appears in Tanzanian school books. In a similar way Iraq is often men­ tioned in the oral traditions of the Iraqw, see Rekdal (1998) for an analysis of the crea­ tion of this myth. Thus, the first lines in an oral history should normally be taken as mythical. For the Mbugu, the place of their ultimate origin that is mentioned most of­ ten is Lukupia, the Mbugu pronunciation of Laikipia, a plateau north of Mt. Kenya; Kericho (Kelicho), which is also mentioned, is a major town in the direction of the Laikipia area. A discussion of the possible relevance of Laikipia as origin for the Mbugu people is taken up in section 2.8. For point 2, movement to South Pare (Vudee or Lasa), nearly all traditions men­ tion Vudee or Lasa as a place where the Mbugu used to live. The accounts do not agree on the place where they were before their arrival in Vudee, if they make state­ ments about this at all. The two main competing views are that they came through

HISTORICAL AND GEOGRAPHICAL BACKGROUND

21

Mbugwe in Northern Tanzania or that they passed through the Maasai plains. These views do not exclude each other; rather the differences lie in which of the elements — 4: Mbugwe or 5: Maasai— are mentioned and in which order. Some of the accounts that mention Mbugwe, point out the similarity in name. The major reason to assume that the similarity in names is no coincidence is the fact that the area where the Mbugwe live is near the Iraqw and Gorwaa, and settlement in such an area would ex­ plain the clear presence of Gorwaa or Iraqw material in Mbugu. I discussed this with Thomas K. Kassi, (January 1994), a local teacher who studies the history of the Mbugwe and according to him the Mbugwe oral history reports only on Rangi, Iraqw and Datoga people entering Mbugwe society. If they, the predecessors of the Mbugu, did not live in that area, then only an influx from Gorwaa or Iraqw people into the Mbugu population can explain the high number of lexical items specifically of Iraqw/Gorwaa origin. In fact, there is a report that some Gorwaa families who had left their homes have been traced down among the Mbugu in the Usambara mountains (Nordbustad, p.c.) but unfortunately I was not able to obtain more information on this. Another intriguing but far from conclusive point is the Inner Mbugu word for 'mother', lagè . If this does indeed originate from the identical Iraqw/Gorwaa denomi­ nation for a cow acquired in war, then there is a suggestion of mixed marriages among those "Mbugu", presumably with Gorwaa/Iraqw men acquiring other women in war. The metaphor cow - woman within an East-African cattle culture is not surprising. For point 3, the dispersal from South Pare, several directions are mentioned: Northern Pare, Maasai and Zigua area (Handeni), and the Usambara mountains. Not all the acounts mention all these directions but all three directions of dispersal are con­ firmed by independent linguistic evidence. Some sources mention a fight with the Maasai about cattle. The prominence given to this fact reflects the importance that the Mbugu see in cattle as one of their prime cultural characteristics. For point 5, a period of serfdom under the Maasai is mentioned by several ac­ counts. That it does not show up in the summaries and analyses by the historians Conte, Feierman, and Kimambo possibly reflects that this detail was not considered vital. On the other hand, the fact that not all the accounts mention a period among the Maasai may very well reflect reality, i.e. probably not all Mbugu clans share this part of the history. For point 6, entering the Usambara mountains is either in the north or in the west. This reflects different migration routes of the various clans. There is full agreement about the six clans that comprise the Mbugu: Nhkangwana, Ombweni or Londo, Nhkandu or Nhkanzu, Ngarito(-Ngarire), Ombeji, Gonja. The order in which they are presented here has some relevance for Mbugu initiation. In addition there are two Mbugu clans that are restricted to the Pare area: Semo and Mrita. The Nango, who are now Shambaa, were once Mbugu; Copland (1933/34) mentions six subclans of the Nango; in addition he gives the name of Dufu for those Mbugu who became integrated into Zigua society. Clan names are given by Storch (1895:324), Copland (1933/34), Green (1963), Feierman (1974), Besha (1974), Kiinzel et al. (1982), Mr. Kakai in Brenzinger (1987) (which contains an overview of these

22

CHAPTER 2

six reports on clan names (Brenzinger 1987:70)), Conte (1994:47), Mr. Sewia, and the elders in Kisiwani (see above). The clan names are presented in Table 2.1. Clan Londo / Ombweni Nhkangwana Nhkandu or Nhkanzu Ombeji Gonja Ngarito-Ngarire

Point of entry Mombo Mombo Shume

Initiation in:

Name of clan founder Vudee SengandaVudee Senganda Suji Suwende 9 Suji Suwende Shume Suji Suwende 9 Suji Suwende Table 2.1: Oral traditions, clans

Present settlement Magamba Bumbuli East of Magamba Rangwi Rangwi 9

In the Rangwi area, the Gonja clan does not speak Inner Mbugu. In the Bumbuli area the Nhkangwana clan is very prominent and they all prefer to speak their dialect of Inner Mbugu. The Lomboza who are recent immigrants to Magila near Bumbuli com­ ing from East of Magamba have a preference for Normal Mbugu. Feierman (1974:81) reports that the Nhkangwana and the Ombweni (Londo) clans went to Vudee for their initiation rituals while the Ombeji, Gonja, Nhkanzu and Ngarito clans went to Suji (both places are in South Pare). One crucial factor in understanding the history of the Mbugu is that not every­ body shares the traditions about a period of submission to the Maasai. The route of settlement in the Usambara mountains is either directly from the South Pare mountains or via the lowlands (Handeni/Zigua, Maasai plains, Ngulu area, and/or Mbugwe). The point of entry into the Usambara mountains was Shume for a large part of the Mbugu (also for the Nango and many subsequent groups), but others came in roughly where the present-day main road enters the mountain area, i.e. Mombo, and their earliest set­ tlement is presumably near Bumbuli. Mr. Ishika in Malibwi (see above) made a dis­ tinction between "us" and "the others". "Us" are the Londo or followers of Ombeni and the Nhkangwana; "the others" include the Nhkandu and the Gonja. The Londo and the Nhkandu used to be together but then a split occurred in Vudee. The Nhkandu and the Gonja followed the Nango into the Usambara mountains; the Londo and the Nhkangwana went to Zigua land first. This division into two kinds of Mbugu clans correlates with the difference in place of initiation that Feierman reported and with the difference in clan forefather that is reported by Kiinzel et al. (1982). It weakly corre­ lates with a preference for Inner Mbugu in the sense that the Nhkangwana in Bumbuli strongly prefer Inner Mbugu and that in the Magamba area the preference for Inner Mbugu diminishes to the east where more Nhkandu live. The difference in proficiency in Inner Mbugu or Normal Mbugu is significant historically because it does not corre­ late with the presence of Pare speakers in the direct environment and thus it cannot be attributed to recent influences. There are clear correlations between clan, point of de­ parture, place of initiation, route of migration, place of entry into the Usambara moun-

HISTORICAL AND GEOGRAPHICAL BACKGROUND

23

tains, time of entry and preference for or knowledge of Inner Mbugu. These correla­ tions cannot be made more precise because in the present-day situation after very in­ tricate movements within the Usambara mountains they are no longer straightforward. A detailed study of clan histories might reveal some of the original correlations. The areas of settlement indicated in Table 2.1 above are approximations; there is no strict clan settlement pattern; people of all clans can be found in every Mbugu area. There are also differences in chronology. It is clear that the first Mbugu to enter the Usambara mountains were the Nango; I assume that the Gonja in Rangwi are the latest arrivals. It has been proposed that this might have been around the beginning of the 19th century (Baumann 1891:183, quoted in Brenzinger 1987:77). By relating the events in the Mbugu oral history to those of other people, specifically the Shambaa and the Pare, rough time estimates have been made. The Nango must have entered the Usambara mountains before the Kilindi dynasty, which leads Brenzinger to estimate that they were there already before 1650 (Brenzinger 1987:75). Kimambo (1969:62) reports that oral traditions agree that the Mbugu and their leader Gurumi arrived twelve generations ago in South Pare; this is four generations earlier than the Kilindi arrival in the Usambara mountains which happened eight generations ago. It is evident that such time estimates are far from definitive and that any episode prior to the South Pare settlement must be placed in mythical times. Before investigating the various lexical source languages and areas in more de­ tail, it is important to recognise that there are — at least — two different Cushitic sources of the Inner Mbugu lexicon. The evidence for two separate Cushitic sources consists of the fact that Inner Mbugu lexemes share certain innovations that sets Iraqw and Gorwaa aside from the other Southern Cushitic languages, see Table 2.2 below. The "other" Cushitic source and Iraqw/Gorwaa are only distantly related. Cognates for Ma'á with the "other" Cushitic source could theoretically be retentions of ProtoCushitic lexicon that were lost in the rest of Southern Cushitic, as Ehret argues in his Table 2 (Ehret 1980: 320-337). This position is difficult to maintain in the case of words, assigned here to the "other" Cushitic source, which are both attested in Eastern Cushitic and in Southern Cushitic, but which formally fit better the Eastern Cushitic etymon. For example, the Inner Mbugu word míndá (9/10) 'house' is in form closer to the Eastern Cushitic cognate mana or mina than to the West-Rift-Southern Cushitic one, mara, see Kießling and Mous (to appear). Within Southern Cushitic it is specifically Iraqw/Gorwaa that is a source for In­ ner Mbugu. The evidence that points specifically to Iraqw/Gorwaa is the following: First, Inner Mbugu shares with Iraqw/Gorwaa the lenition of West-Rift voiced stops, d>r, b>w, g>y, cf. Ehret (1980:87) and Elderkin (1988:84), as we can see in Table 2.2 lu-haremú 'horn', we 'buy' and possibly also in m-he 'person' and 'a' 'eat', although these last two examples are not conclusive because the truncation rule could have been in operation, see 3.4. Secondly, the vowel e in lu-haremú and y-asemú points to Iraqw/Gorwaa as the source because proto West-Rift had a which Iraqw/Gorwaa inno­ vated to e. Thirdly, the lateral fricative in i-'ohlu corresponds to a development from ejective affricate to lateral fricative typical for the transition of West-Rift to

24

CHAPTER 2

Iraqw/Gorwaa. The remaining examples in Table 2.2 are Iraqw/Gorwaa lexical innova­ tions. A more detailed treatment of this issue can be found in Mous (1996). Gorwaa is more likely than Iraqw to be the source for Inner Mbugu, see 2.5. The two languages, Iraqw and Gorwaa, are closely related.4 lu-haremú also ln-xaremú 'hom' from Iraqw(/Gorwaa) xareemi (pi), cf. West Rift SC *xadamo. wé (v) 'buy' from Iraqw(/Gorwaa) weer-iis 'to sell' (containing causative); cf. West Rift SC *beer-iis. 'a (v) 'eat' from Iraqw(/Gorwaa) by; cf. West Rift SC * ag. m-hé 'person' from Iraqw(/Gorwaa) hee; cf.West Rift SC *hiid-. y-asemú 'breast (female)' from Iraqw(/Gorwaa) iseemo (pi); cf. West Rift SC *isamo. i-'ohlú 'cheek' from Iraqw(/Gorwaa) unhl; cf. West Rift SC * untla. 'ihlé (aj) 'unripe, bad' from Iraqw(/Gorwaa) unhl 'fresh' (or from Yaaku 'ihle' 'pasture, grasses'); there are no other West Rift SC cognates. búre 'short rainy season' from Iraqw(/Gorwaa) buhare 'rainy season'; there are no other West Rift SC cognates. m-mú (Bumbuli variety) 'person' from Iraqw(/Gorwaa) mm 'people'; there are no other West Rift SC cognates. sa'á 'wind, cold' from Iraqw(/Gorwaa) tsa' 'be cool'; there are no other West Rift SC cognates. m-lagé 'woman, wife' from Iraqw(/Gorwaa) lage which is a cow-name for a cow acquired in a war, also in Mbugwe; there are no other West Rift SC cognates. Table 2.2: Ma 'á words that share innovations with Iraqw/Gorwaa. In the following section we have a closer look at the origin of the Ma'á parallel lexi­ con, the distribution of the lexical sources and the possibility of drawing historical conclusions from these. 2.2

Historical information from th e parallel lexicon The basic properties of the parallel lexicon were outlined in 1.7. There I argued that Ma'á is in fact an extensive parallel lexicon to Normal Mbugu in which the Inner Mbugu (Ma'á) words are on a par with Normal Mbugu words. However, not all lexi­ cal entries occur in pairs. An important number of lexical items are shared by Inner and Normal Mbugu. This is especially true for specialised vocabulary and recent loans. Many of these shared lexical items originate from Shambaa or Swahili, see 2.9. In addition, there is a gradual expansion of Normal Mbugu lexicon at the expense of Inner Mbugu. Younger people use the words in the IMb/NMb column of Table 2.3 below for both languages, whereas older people still remember the archaic equivalent Inner Mbugu form. Another example, is hongé(r)a which is the most common expres­ sion for 'thank you' and a Shambaa term. There is a "real" Inner Mbugu word háyka

HISTORICAL AND GEOGRAPHICAL BACKGROUND

25

parallel to this normal term. Older people pointed out to me that in fact there once was yet another "more original" Inner Mbugu term, yahálo, which is no longer used. IMb-archaic gali nkhútó kuhl kumumú gimbé kizógoa sugháno hima

meaning to touch hyena long, high short, narrow noise shoulder right (?) hand left (?) hand

IMb/NMb doa i(n)gúgwa ezá sisírí kilóndó kitúro kuume kumosó

Table 2.3: Archaic Inner Mbugu words The diachronic aspects of the parallel lexicon are also visible in changes in pairings of words. Such changes are evident from the dialectal variation within Mbugu; for exam­ ple the verb 'to hunt', is kasha in Normal Mbugu and kalá in the Inner Mbugu of the Magamba area but hlára in the Inner Mbugu of the Bumbuli area. The origin of the verb ku-hlára is unclear; ku-kasha is Pare, ku-kála is probably from Zigua ku-kala 'to hunt' or possibly from Shambaa u-kalá 'hunting'.5 In Inner Mbugu of the Bumbuli area, people distinguish between 'cow blood', saxó, and 'other blood', háha. The word for 'cow blood' is what the people in the Magamba area use in Inner Mbugu for any blood, sakó. The Bumbuli word for 'other blood' is the word that the people in the Fwizay-Rangwi area use in Inner Mbugu for any 'blood', háha. The Normal Mbugu word is sakame (from Pare, and ultimately from Cushitic) in all areas with no distinc­ tion. Cow blood used to be drunk mixed with milk, as is common among most cattlekeeping people of East-Africa. Other such cases are the words for 'cow, cattle'. In the Bumbuli area, dé and wá are a suppletive singular/plural pair for 'cow, cattle', reminis­ cent of hiee (sg), hikwa (pl) 'cow, cattle' in Iraqw. In the Magamba area, however, there is no suppletive singular/plural pair for cow. There is one word, wá (9/10) 'cow, cattle'; and the word dé is used for 'goat', and also for 'calf. Meinhof (1906) and Green (1958) recorded de meaning 'oxen', but Storch (1895) has nde or wa for 'cow'. The differ­ ences in lexical structure possibly reflect a higher degree of influence from an Iraqw/Gorwaa lexical pattern. If so, we have to assume that the Iraqw/Gorwaa influ­ ence is stronger in the Bumbuli area. Nowadays 'rainy season' is bure, but in 1906, in Meinhof's days, it was buxéta (Meinhof 1906), and in Green (1958) there is a distinc­ tion ibuhéta 'long rains' versus bure 'short rains'. This lexical distinction has since been lost. Some historically important observations on the make-up of the vocabulary are the following:

26

CHAPTER 2 1. Words that have no Inner Mbugu parallel are Bantu: Pare, Shambaa or Swa­ hili. New loans from Shambaa or Swahili are often not parallel. 2. Normal Mbugu words are from Pare if they have an Inner Mbugu parallel. 3. Forms that can be shown not to have come from Pare (or Shambaa, or Swa­ hili) belong to Inner Mbugu and have a Normal Mbugu equivalent. 4. Words that contain a velar voiceless fricative, a lateral voiceless fricative, or a glottal stop belong to the Inner Mbugu variety and have a Normal Mbugu equivalent.

The characteristics of the lexical structure in Inner Mbugu are different from usual borrowing. These are the observed parallelism (cf. section 1.7), the fact that basic vo­ cabulary is affected more than specialised vocabulary and the relatively high propor­ tion of verbs. Usually the borrowing of verbs is rare when compared to that of nouns and if it does occur, the form often needs a verbalising derivation from the receiving language. There are no signs of obligatory verbalising morphology in Inner Mbugu and the number of verbs in 1Mb is considerable. Figure 1 gives the breakdown of the Inner Mbugu lexicon according to parts of speech, and at the same time the various sources of the Inner Mbugu lexicon which shows that in particular in words that origi­ nate from Maasai the proportion of verbs is remarkably high, even larger than that of nouns. This fact is possibly of significance for the reconstruction of the linguistic his­ tory of Mbugu.

Figure 1: Word category and origin in Inner Mbugu The various sources for the extra Inner Mbugu lexicon can be divided up as in Table 2.4 below, namely words that are adapted from Normal Mbugu; pre-Iraqw are those words that are closely related to Gorwaa/Iraqw (IR). The comparison is with Iraqw rather than Gorwaa because our data on Gorwaa are limited compared to those on Iraqw. "Other Cushitic" are those words that are Cushitic but not from the Gor­ waa/Iraqw source, that is, corresponding to Oromo (OR) or Dahalo. Here we often use Oromo as point of reference, not to suggest that Oromo was the donor language of

HISTORICAL AND GEOGRAPHICAL BACKGROUND

27

these roots but rather because, again, lexical research on Oromo is more advanced. I use the terms "pre-Iraqw" and "other Cushitic" as labels to make a distinction between the two separate Cushitic sources of lexical material in Inner Mbugu, see also Mous (1996) and Table 2.2. Other words are from Maasai (MA), from the Taita Bantu lan­ guage Sagala (SA). Taita Bantu refers to the languages Sagala (SA), and its neighbour Davidą, both spoken across the border in Kenya in the Taita mountains. Table 2.4 shows that all these sources are relevant for the Inner Mbugu lexicon. Figure 2 pre­ sents the relative importance of these various sources. 1Mb lu-xaremú wé m-he m-mú (Bum) fwaru aro hará i-tiráo i-diyé Iaha nyamâlo zéu i-hlarí raó hla

meaning horn buv person person sing elephant (archaic) calf liver dog loose work herd cloud arrow for bleeding cattle ache

origin IR: xareemi (pl); WR: xadamo IR: weer-iis 'to sell'; WR: beeriis IR: hee; WR: hiidIR: mnu 'people'; no WR OR: jaaruu 'song, poem' OR: arba OR: halaa 'young female camel' OR: tiruu MA: ol-die (in Mbugwe diyo) MA: a-laa MA: a-nyamál 'be busy' MA: a-reu 'drive cattle' SA: tlyara SA: ki-rao 'barb of an arrow' SA: lya

Table 2.4: Diversity of origins of Inner Mbugu words

Figure 2: The distribution over source languages for the Inner Mbugu lexicon In the following paragraphs we trace back the route from the oral history going back

28

CHAPTER 2

in time, discussing other aspects of these oral traditions and presenting the etymologi­ cal evidence from the five relevant areas: Usambara mountains, Maasai, the Mbugwe Iraqw area, Pare mountains, and the North, that is, the Taita mountains and the other Cushitic source. 2.3

Usambara mountains: Shambaa, Seuta Bantu, Swahili All the Mbugu who speak Inner Mbugu (Ma'á) live in the Usambara mountains where they live among the Shambaa, the dominant population in Usambara. There are also a considerable number of Pare people in Usambara. Every Mbugu has Shambaa as neighbours and nearly every Mbugu also speaks Shambaa. In recent history the Mbugu lived in areas in Usambara where they had little contact with the Shambaa. They lived with their cattle in the higher areas which were not used by the Shambaa for cultivation (Conte 1994:48, Hailey 1950:327). New (1875:419) describes the Mbugu as "not unlike degenerate Wakuavi [Maasai], who, driven to Usambara for refuge by their foes, have, in the course of time, lost their language, and become as­ similated to the people with whom they are living, in everything but an unalterable physique and a few incorrigible manners and customs. The Wasambara [Shambaa] are just what you expect to find the aborigines of such a country to be." But in those times too there was contact with the Shambaa and both the Shambaa and the Mbugu oral histories make mention of each other. The situation was different for the Nango. The Nango are a people of Mbugu descent who are no longer Mbugu in the eyes of the (other) Mbugu because they have taken over the Shambaa way of life, that is, their language, customs, and economy. Whereas for the rest of the Mbugu, the relation with the Shambaa can be described as living apart together, this certainly does not hold for the Nango. The Nango are not fully accepted by the Shambaa, but the relation with the Shambaa was always an important issue for both the Nango and the Shambaa. At various points in history alliances were made or reconfirmed between the two groups, for example by Kilindi rulers marrying Nango wives (Feierman 1974:93) and the Nango thus endorsing the hegemony of the Kilindi clan. The Nango were the first to enter Usambara and they arrived at least a century before the coming of Mbegha, the legendary founder of the Kilindi dynasty, see Feierman (1974:61 ff). Feierman puts a date to the reign of the Kilindi Kimweri ya Nyumbai which started in approximately 1815 (Feiererman 1974: 93nt4; 101 nt l 2) and he reports agreement in the oral tradi­ tions that the Kilindi reign started four generations earlier, i.e. roughly in the early 18th century. The Nango were originally part of the Gonja in South Pare (Feierman 1974:76). Nango kept their separate initiation rite in Shume where they all gathered (Feierman 1974:81). Furthermore, they were politically divided and some of them pre­ ferred Kilindi power to their own (Feierman 1974:89). The Nango were eventually defeated and integrated by the Shambaa (Feierman 1974: 90). It is likely that ultimately a large portion of the Shambaa population in Usam­ bara arrived from Ngulu, (Feierman 1974). This fits with the linguistic classification of Shambaa in a Seuta Bantu group with the Bondei (East Usambara), the Zigua (southern neighbours in the lowlands) and the Ngulu (further to the south-west). This

HISTORICAL AND GEOGRAPHICAL BACKGROUND

29

group of languages is named after a legendary hunter hero Seuta (se is a naming pre­ fix, uta 'bow') who is known to all these people. The words that Mbugu has borrowed from Shambaa are mainly new words that are shared in both languages. In addition, Pare words in Normal Mbugu are replaced by or adjusted to Shambaa to some extent. A Shambaa word is never used as the Inner Mbugu equivalent of a Normal Mbugu or Pare lexeme. Shambaa, which is the domi­ nant language in the area, is too close to Pare to be a source for Inner Mbugu words since the purpose of these is to be deviant from a Bantu language. Table 2.5 gives some, though not all, of the loan words that can be shown to be Shambaa (Sh) rather than Pare (Pa). In the etymological wordlist (see Appendix) many more words in Normal Mbugu have a Shambaa source, but this is partly because the lexical data on Shambaa are much richer than those on Pare and a corresponding lexeme with Sham­ baa could be found while for Pare the data were insufficient. NMb word ki-ogha kimba i-lua

pucha (v) nhkúngúní shí m-nyáu lu-sayâ

meaning mushroom corpse flower to pluck bedbug land cat cheek

language I+NMb I+NMb I+NMb I+NMb I+NMb NMb NMb NMb

source Sh: ki-ogha (Pa: ki-choghd) Sh: kimba (Pa: kiimba) Sh lua (Pa: i-ud) Sh: pucha (Pa:fiithà) Sh: nhkunguni (South-Pa: nkungwini) Sh: shi (Pa: thí, sí) Sh: m-nyau (Pa: i-nyáwi) Sh: lu-saya 'row of teeth' (Pa: ithä)

Table 2.5: Some Shambaa words in (Normal) Mbugu Zigua (bantu) is a candidate tor being a source ot the (Inner) Mbugu lexicon based on the oral tradition of going through Handeni on the way to the Usambaras; and there was also an influx of people from Zigua area into the Usambara area. In fact, very close to the Mbugu village near Bumbuli, there is a village of people originating from Zigua country. The Mbugu must have been in closer contact with the Zigua when they shared the Maasai plains with them. In the 19th century the Zigua must have been dif­ ferent from the Shambaa in economical terms and more similar to the Mbugu since both Johnston (1879:553) and New (1875:416) describe the Zigua as a pastoral peo­ ple. During my brief stay in Handeni in the Zigua area I did not find indications of remnants of a Mbugu population, nor evidence for knowledge among the Zigua people about historical links with the Mbugu. There is no Zigua influence on the Ma'á. There are a handful of words in Inner Mbugu that have been claimed to be of Zigua origin, see Table 2.6. But for all of these words there are competing solutions as to their ety­ mological source.

30

CHAPTER 2

NMb gugúlu

meaning run

language 1Mb

gimbé

noise

makalé

charcoal

IMbArchaic 1Mb

kalá

hunt

1Mb

source Zigua: guluka, or Oromo: gugad'd'a 'run away in fear'. Ehret (1980) has Dahalo: gírigíri 'fast' Zigua: gamba 'talk', also present in Alagwa and Burunge (both South Cushitic) gimba 'talk' Zigua or Mbugwe: makâlą plus addition of éno and subse­ quent truncation Zigua: kala. Shambaa does not have this verb but probably once had it given the presence of the noun ukala.

Table 2.6: Inner Mbugu words ofpossibly Zigua origin6 Ehret (1974) posits Mbuguan (predecessors of the present day Mbugu) presence in the Zigua area around two thousand years ago, and on the basis of evidence of four Mbugu loans in Zigua-Shambaa he concludes that most of these Mbuguans were ab­ sorbed by the Zigua and Shambaa. Ehret's four loans are the following (Ehret 1974: 69, see also Brenzinger 1987:40ft27): 1. Proto Shambaa-Zigua -hoya 'axe' from Ma'á hoyu (I have 1Mb ihoya). The only Southern Cushitic link is Asa hawo 'axe' (Ehret 1980:307). The other direction, Zigua to Inner Mbugu and Asa is more likely since it is the regular reflex of ProtoBantu *-oga 'hoe, axe' (Coupez et al. 1998). 2. Bondei (Seuta Bantu) bomo 'crocodile' from Ma'á boma 'snake' which is, accord­ ing to Ehret (1974:69; 1980:138), ultimately derived from proto-Southern Cushitic *bom- 'to lie, to sleep'. The proposed link is Inner Mbugu bodi 'lie down' in which the first syllable can be related to Dahalo  'lie' (the second syllable is di 'down') and Oromo-Waata bofa 'to lie'; while these last cognates may be ac­ cepted, the semantic connection between 'snake' and 'lie down' seems particu­ larly dubious; moreover, it is unnecessary since -borna 'python' is reconstructed for proto-Bantu (Meeussen 1969). Thus the borrowing is from Bantu into Inner Mbugu rather than the reverse. 3. Zigua nofi 'weaverbird' from Ma'á i-no 'ofi (6). I have i-nofi (5) for both Inner and Normal Mbugu and, as is the case with most birds, this is a borrowing from Shambaa nofi (5) 'weaverbird'. Ehret does not present evidence for a Cushitic source other than the Ma'á word itself. 4. Zigua gwirizi, or Bondei ngurizi 'eagle' from Ma'á gwirizi and ultimately from proto-Soutern Cushitic *gwariz-. The word, ngwirizi is found in both Inner Mbugu and Normal Mbugu as is the case with most specialised flora and fauna. More­ over, there is a possible Shambaa source, ngwilizi 'eagle' (LangHeinrich 1921:367) —the variation / or r is dialectal in Shambaa. For the lexical items that are in the common lexicon I assume the Pare or Shambaa equivalents as source and thus we can assume this lexical item to be a recent loan into (I+N) Mbugu. This is not to exclude the possibility of Cushitic as an ultimate source for the ety­ mon in Shambaa and other Bantu languages, for example Pare has ngwirini for

HISTORICAL AND GEOGRAPHICAL BACKGROUND

31

'eagle'. What is interesting is that the Seuta languages and Pare have an r which is also present in Ehret's construction of Southern Cushitic. Kießling and Mous (to appear), however, have *gwadiisi 'vulture' for Proto West-Rift Southern Cushitic and Alagwa, Gorwaa and Iraqw lenited the d (which is preserved in Burunge) to r. Given the time depth, it is difficult to imagine that Seuta Bantu and Pare borrowed the word from a Southern Cushitic source after Burunge split off from the rest of West Rift Southern Cushitic. Furthermore, if the etymon is ultimately from a Cushitic source, it has spread over the entire area and it does not provide specific evidence for early Mbugu presence in Zigua area. Concluding we can state that there is no evidence for Mbuguan influence on Zigua or Proto Zigua-Shambaa. The fact that there is no lexical influence from Shambaa and Zigua on Ma'á, two languages with which the people are and have been in close con­ tact, shows that the Ma'á or Inner Mbugu words have to be deviant from Normal Mbugu (Bantu) and Shambaa and Zigua are considered to be too close. 2.4

Maasai According to the oral traditions, the Mbugu, or part of the people now called Mbugu, lived as a group among the Maasai doing certain herding activities for them. The early reports on the Mbugu also depict them as Kwavi, (New 1875:419), i.e. as being of Maasai origin. There are quite a few words in the Inner Mbugu lexicon from Maasai and among them such words as 'work' and 'to herd'. This linguistic evidence supports the historical interpretation of the claim in the oral history of a longer period of living among the Maasai. Moreover, it is common to have separate subservient groups called "Dorobo" living among the Maasai. That such groups do herding tasks is less common but not unheard of. Kassam (1986:197), citing Torry (1973:59), re­ ports for the Waata (Oromo), a servant "Dorobo" group among the Gabbra Oromo, that one of the menial tasks they do is herding small stock, and Heine (1975:31) re­ ports such a "Dorobo" group that was herding goats for the Samburu (Nilotic). The groups that I encountered in the Maasai plains were quite heterogeneous as regards their individual linguistic background. It was there that I was searching, alas in vain, for the Aasáx (or Aramanik) who were just such a group speaking a Southern Cushitic language and on which we have very little information. It remains uncertain whether or not they played a role in Mbugu history. The Tanzanian dialect of Maasai that most likely served as the Maasai source for Mbugu is Parakuyo, but in the absence of reli­ able data we have to resort to grammatical and lexical descriptions of the Kenyan dia­ lects. The Parakuyo came from an area north or north-west of the Laikipia plateau claiming Kerio as their place of origin, and by the turn of the eighteenth and the nine­ teenth century they were expelled from this area and eventually settled in the Pangani river area which is at the foot of the Usambara mountains, see Sommer and Vossen (1993:34-35). A number of Maasai lexical items that contain an r in Maasai have a z instead in Ma'á. There is no evidence for a sound change z > r in the history of Maasai and r is a

32

CHAPTER 2

common sound in Mbugu, thus this change remains a puzzle to me. Some examples of Maasai words in Inner Mbugu are given in Table 2.7. Inner Mbugu oké 'because' arjâ 'home' ln-'ar åme 'broad' m-sá 'body' ny am alo 'work' pu'ú (v) 'shine, appear (of sun)' púru (v) 'jump, fly, snatch' ré (v) 'return (tr)' siki (v) 'dust, rub off sú (v) 'pour' zéa (v) 'herd' ii (V) 'forset'

Maasai ake 'only, just' ân, enk-an) 'home' a-lala 'be wide' plus suffix -mé o-sesen 'body' a-nyamál 'be busy', e-nyamáli 'trouble' a-piup 'blaze (of fire)' a-ipirri 'flee, fly' (or onomatopoeic) a-rriny-oa 'return there' a-sik 'scrape off hide, rub, polish' a-isuak 'pour down, spread (saliva)' a-rew 'drive cattle' (or from Iraqw/Gorwaa de'-eem) a-ri-ki 'forset'

Table 2.7: Inner Mbugu words of Maasai origin Gorwaa and Mbugwe The name Mbugu is practically identical to that of Mbugwe. The Mbugwe are a Bantu people who live between the Maasai and the Iraqw/Gorwaa.7 The correspon­ dence gu = gwe is common in Pare. These people speak a language that is closely re­ lated to Rangi (Bantu), which is spoken around the town of Kondoa, further to the south. Between the Mbugwe and the Rangi (=Langi, Bantu) live the Gorwaa and the Alagwa, two Southern Cushitic groups. The Iraqw are neighbours to the Mbugwe but there is a large vertical distance between the two, i.e. the depression of the East Afri­ can Rift Valley. According to some oral traditions the Mbugwe originated from Rangi companies searching for salt.8 At the same time the Mbugwe are known to have been rich in cattle once, and, living in the open plains, they were an easy target for the Maasai. The question is: Is the reason that the Mbugwe are mentioned in some of the oral histories of the Mbugu solely to explain the similarity in name or is there a genu­ ine link between these two people who carry the same name? The possibility of a ge­ netic linguistic link between Mbugwe and Pare remains to be researched. Nurse (1979:36If) draws attention to lexical similarities between Rangi (and Mbugwe) and especially Pare while they belong to different groups on the basis of phonological isoglosses. There is virtually no linguistic material in Inner Mbugu that points to Mbugwe. I could find only three items and two of the three have alternative sources: mringa (3/4) 'bee hive' [IMb(Rangwi)] (NMb: mzinga) from Mbugwe mringa. Other Bantu languages in the area do not have r in this root; possibly idiyé (5/6) 'dog' [1Mb] (NMb: iguro) from Mbugwe -dia, ultimately or directly a loan, also in Digo, from Maasai ol-die; makalè (6) 'charcoal' [1Mb] (NMb: makaa) from Mbugwe or Zigua makála (see Table 2.6), or from Normal Mbugu makaa before /-loss.

2.5

HISTORICAL AND GEOGRAPHICAL BACKGROUND

33

The Gorwaa are the Cushitic people closest to Mbugwe. The lexical evidence in Inner Mbugu points to Gorwaa or Iraqw. The two languages are very closely related but the Gorwaa are in closer contact with the Mbugwe. I therefore assume that the Southern Cushitic source for Inner Mbugu lexicon was Gorwa rather then Iraqw. Moreover, some Gorwaa remember that a Gorwaa family went to live with Mbugu in the Usambara mountains only a few decades ago (Nordbustad p.c.). 2.6

Pare The Pare mountains consist of three mountain ranges. The South Pare moun­ tains are very close to the Usambara mountains, at a visual distance, and people, in­ cluding the Mbugu, move and have been moving between the two areas for some time. The Mbugu also have a connection with North Pare. It is in North Pare where people live who are known to be of Mbugu origin, near the place Mbughweni (Mbugu-LOC). This is also the area where the Gweno live. Gweno is a Chaga (Bantu) language that is converging with Pare (see Mreta 2000:167ff). There are no remnants of the Mbugu language in the Pare area. There are no Mbugu words in ceremonial texts, not even in songs, and nobody remembers their grandparents as having spoken Mbugu. Some people in the Pare area are aware of their Mbugu roots and of the fact that they have their brothers in Usambara and that these have their own language, which they assume to be the original language. According to some oral traditions, the Mbugu in North Pare and those in Usambara both originate from South Pare. Kimambo (1969:62) also found that Pare and Mbugu traditions agree that the Mbugu arrived later than the Pare in the North Pare mountains and that they moved from South Pare to North Pare, and that a large group of Mbugu went to Usambara from South Pare. The genealogies in­ dicate an arrival of the Mbugu in South Pare 'about twelve generations ago. The Mbugu people always possessed more cattle than the Pare people. The claim of the oral traditions that the dispersal area was in South Pare is substantiated by linguistic facts in the sense that Normal Mbugu is predominantly South Pare, see Table 2.8. Normal Mbugu kinhtu 'thing' nhkíngo 'hide' safu 'ant (bitinę)'

South Pare kinhtu (also Shambaa) nkingo 'skin, hide' sáfu

North Pare kindu ngingo tháshu

Table 2.8: Some lexical evidence for the South Pare affiliation of Normal Mbugu (small excerpt) Substantial dialectal differences exist between North and South Pare. North Pare has voiced all nasal-stop clusters (but not South Pare, nor Normal Mbugu) and it has den­ tal fricatives where South Pare and Normal Mbugu have alveolar ones. Lexically, however, Mbugu is sometimes closer to North Pare: dindika 'run' from North Pare díndíka rather than South Pare jinka, and gonda 'carve, write' from North Pare gonda rather than South Pare or Shambaa gonta.

34

CHAPTER 2

Among the Pare dialects, North Pare is deviant as a result of Gweno influence. The Gweno influence on Mbugu is difficult to trace, since there is very little material on Gweno, see Philippson and Nurse (2000) and Mreta (2000). The few items that can possibly be traced back to Gweno are Inner Mbugu (Ma'á) while the lexical items that are from North Pare rather than South Pare are in Normal Mbugu. This suggests that these two influences are to be taken separately. The North Pare influence is probably due to movements between the Pare and the Usambara mountains. The Inner Mbugu words that can be traced down to Gweno are: mirá (adv) 'but' from Gweno mirá, diláo (la/2) 'king' from Gweno mlaú 'headman', rather than from Oromo d'iira 'man, men, husband(s), male(s)', or Qwadza (Southern Cushitic) dilu 'older (in kinship terms)' (Ehret 1980), and iréno (5/6) 'banana leaf from Gweno iru 'banana leaf, with the addition of an ending ~eno. The evidence for Gweno in Inner Mbugu is meagre and not corroborated by other links between Mbugu and Gweno. It is probably a relic from the time that the Mbugu were among the Pare. Inhabitants of the Pare mountains had continuous contact with those in the Taita mountains in Kenya. This contact consisted of a population influx from Taita, for trade, specifically ivory, marriage —also at the royal level—, and war. The influx of four clans in South Pare from Taita about ten generations ago gave rise to a new po­ litical order and unity in that area. One of the results was the elimination of the previ­ ous non-Pare inhabitants of the area, the so-called Wagalla. Kimambo (1969:28f) ar­ gues that these Wa-Galla are different from the Oromo (also called Galla); he argues that various unrelated groups are referred to by the name of Galla. He does acknowl­ edge that there is evidence for early non-Pare settlers in South Pare that have been ab­ sorbed by the Pare. There is no addtional indication to link the Mbugu to these "Galla". Some additional points that the Pare and the Mbugu have in common are that both are highland people (O'Barr 1973:52), and that they share(d) burial traditions in which the skull is placed in a shrine (O'Barr 1973:51). The South Pare initiation was an open ceremony and not ethnicity- or unity-defining as it was in North Pare and for the Nango (Kimambo 1969:69-71). The Mbugu initiations used to be in various places in South Pare and were possibly shared with the Pare people in South Pare. Finally, the fact that the Bantu side of Mbugu is basically Pare begs the question of the origin of Pare which is in fact unresolved: Pare is a challenge to the subclassifi­ cation of Eastern Bantu and it is set up as a group on its own in classifications by Nurse and Philippson (1975). There are historical links between Pare and Taita. 2.7

The Taita connection In the Taita Hills, two related Bantu languages are spoken: Sag(h)ala and Davidą.9 Davidą is the main language that is spoken on all three mountains: the Taita ridge, Sagala mountain and Kasigau mountain. Sagala is only spoken on (part of) Sa­ gala mountain and is closer to Mijikenda, which is the origin of an important influx of people in the Teri valley of Sagala mountain. It is also the area where the influence of former inhabitants, a people called the Wa-mbisha (or Bisha) is strongest. It is here too

HISTORICAL AND GEOGRAPHICAL BACKGROUND

35

that the first influx of the present-day people, the clan of the Wanya, lived. The dialect of Davidą that has lateral fricatives according to Merrit (1975) is in an area where these Wanya moved to. There are oral traditions of movement between the Taita and the Pare mountains (Kimambo 1969, Merrit 1975) and also between the Taita and the Usambara mountains (Feierman 1974, Merrit 1975). In fact at the beginning of the 20th century there was a small group of Taita people living near Bwiti at the foot of the Usambara mountains who still spoke Davida (Woodward 1913/14) and who had been there since the middle of the 18th century. There is historical evidence for two groups of earlier occupants of Taita: The Wasi, hunters and gatherers, and the Bisha, agricultural-pastoral ists, see Merrit (1975) and Liszka (1974). The term W-asi is a general Bantu word referring to original inhabitants, see Nurse (1979:390-392). Many Bantu peoples have traditions of original habitants called w-asi. In 1932 there was still a small village of so-called Langulu near Sagala (Merrit 1975). The word Langulu, like Wasi and Dorobo, refers to any small group of hunter-gatherers related to cattle nomads, in this case the Oromo. The Bisha are associated with cattle and agriculture in the oral traditions, and stone burials, carved boards and certain iron-working sites are attributed to them. They disguised or gave up their identity. Incursions to the Taita Hills came from north east Kenya, first by the Giriama, and later also by the Kamba, the Chaga, the Kikuyu, the Shambaa, the Taveta, the Pare, the Mijikenda, as well as by the Maasai and the Oromo who were living nearby. There is possibly phonological evidence of Cushitic influence in Taita Bantu in the form of a lateral fricative consonant. Wray (1894) used a trigraph tly in Sagala which, according to Ehret and Nurse (1981:145), is to be interpreted as a lateral frica­ tive, but which now is an implosive (palatalised?) voiced velar stop written as g (Philippson p.c.). In Teri-Davida, according to Harris (1978), "[t]he principal distinc­ tive feature of the Mbale dialect was (is) the occurrence of a voiced lateral fricative in positions where other dialects have a voiced alveolar fricative...". This is also attested by Williamson (1943), who writes 13.10 The words in Inner Mbugu that are related to Taita lexemes are given in Table 2.9. Some contain this lateral fricative. 1Mb hási i-hlarí bughu kuhléa rao xwa'ú m-pahé xará ka'é hlohlóshi pa'a

meaning veins cloud fear ant bleeding arrow last night sugarcane be bitter precede begin cut, saw

Sagala, Davida (Dav) hasi tlyara huga tlyeha ki-rao11 i-hwayn ipane harar-a hay-a tlyoy-an paia

meaning vein, muscle cloud, fog affliction white ants barb of arrow last night bundle of sugarcane be bitter go first begin split

36

CHAPTER 2

1Mb bo'i ogú sáru shighatí bari i-har áo hlá butu hlá t ara'i dúmu hlamí

fi

kuhle laha haí gali kum lire sa m-'lis e ma-gal é it ara

meaning Sagala, Davidą (Dav) boesa do, make jigau castrated goat tsarua14 cool down outside shigati15 go on a trip barita calabash-half ma-hado tlaia stab put into Dav: buduka ache lya shake, curdle milk taras hu16 want, agree durnus a pay shana finish sila long lele be lost lagaia no haiya try gal-usa all, many kamar e like sa ewe that has not lambed Dav: os i maize Dav: m-gere python Dav: are

meaning repair he-goat be content, cool outside take a walk calabash seeds stab fall into ache butter allow pay finish long be lost no change all, whole about ewe lamb grain stalk python

Table 2.9: Inner Mbugu words shared with Taita (Sagala) On the basis of the many non-Bantu lexical elements in Sagala and Davidą, Ehret and Nurse (1981) posit two extinct Southern Cushitic languages, Taita A and B, in addi­ tion to borrowings from Ma'á into these languages. This direction of borrowing from Ma'á to Taita Bantu is based on the fact that these roots are reconstructed for ProtoSouthern Cushitic, (Ehret 1980). I do not accept all the etymologies presented in Ehret and Nurse (1981) and Ehret (1980) but enough remain, or can be added, to prove a link between Ma'á (Inner Mbugu) and Taita. Moreover, a Cushitic influence on Taita is indisputable on the basis of lexical evidence. On the basis of the deviance with regu­ lar correspondences of Sagala within Bantu, many words and in particular those that contain x (which varies with h) or tly are candidates for loans in Sagala. The exact identity and nature of this Taita link with Inner Mbugu is problematic. One possibility is borrowing from Sagala into Inner Mbugu parallel to what we have assumed for the other sources thus far and there is enough evidence for contact between both the Pare and Usambara mountains and the Taita hills that would make borrowing a possibility. Another option is that the Taita words in Inner Mbugu are remnants of an Old Kenyan Cushitic presence represented by the the former Bisha inhabitants of Taita.

HISTORICAL AND GEOGRAPHICAL BACKGROUND

37

In fact, in the direct vicinity of the Mbugu area (between the Usambara moun­ tains and the coast) there are several other groups who, while their identity leaves some question marks, are linked to a more northern area, either on the coast, or in Central Kenya. These are the Segeju and Daisu, see Nurse (1982, 2000),and the Degere and Vuna, see Walsh (1992/93). The latter two are (were?) groups of former hunter-gatherers or elephant hunters. Walsh reports abouth the Degere, that some have retained elements of their earlier speech (1990:69,79), one of them being sako 'girl yet to bear children" 7. Walsh also reports on a group of aboriginal (pre-Waata) huntergatherers who integrated into Giriama society (the origin of the main early Sagala in­ vasion18 in the Taita hills) and who are called Laa (or Alaa). This root also appears in (A)langulu, (A)ryangulo, Mijikenda names for the Waata hunter-gatherers and ele­ phant hunters (Walsh 1990:70-71). Krapf (1858) also used this name, Ala, for a group of hunter-gatherers living between the Usambara and South Pare mountains. Johnston (1886:403) uses the same name for the Mbugu, but Brenzinger (1987:9) argues that Johnston must have erred and relates these Ala to the Aasáx, following Winter (1979:92), and contra Nurse (1979:391) who links these Ala to the Alagwa. I assume that Johnston was referring to Krapf s Ala mentioned above. 2.8

Old Kenyan Cushitic: Eastern Cushitic, Dahalo In this section we investigate the most problematic and probably the oldest lexi­ cal layer in Inner Mbugu: the "other" Cushitic source in Inner Mbugu; that is, those words that are Cushitic but not West Rift Southern Cushitic. The "other" Cushitic source is Eastern Cushitic. Table 2.10 presents some of these cognates for Inner Mbugu. The Mbugu words in the table are taken from Inner Mbugu only, and not from Normal Mbugu, although this variant too contains some (Eastern-)Cushitic loans; for example, kuri 'dog' is Cushitic and Normal Mbugu, but not Inner Mbugu, and it is not of primary concern to us here. For comparison with the "other" Cushitic source, I used Kenyan Borana Oromo from Stroomer (1995), or alternatively Ethiopian Oromo from Gamta (1989) and Gragg (1982). Only the latter two are indicated as such, Stroomer (1995) being my default source. I chose Borana Oromo as representative of the "other" Cushitic source because it is the geographically most proximal Eastern Cushitic language that has been well-documented. 1Mb fwaru (v) warè m-harega áro kå fufa (v) hará

meaning sing kind of hyena arm kind of animal (elephant iri Farler(1885)) get up strangle calf

cognates in Promo faaruu 'song, poem' warabeesa 'hyena' harka 'arm' arba 'elephant' kaa 'stand up, rise' fufa 'add, extend, tie together' halaa 'young female camel'

38

CHAPTER 2

1Mb i-laméto i-tiráo i-hutáo koré (m-)haráza

meaning broom liver hoe pot river

ráru (v) dímu (v)

predict, divine descend

cognates in Oromo lammat 'kind of stick (Gragg) tiruu 'liver' hordaa 'digging stick, kind of dibble' (Gamta) okotee 'pot' hunrsaa 'waterfall' from hursn 'roar e.g., like water in waterfall' (Gamta, Gragg) raagaa 'fortune-teller' d'immuu 'depth'

Table 2.10: Oromo cognates in Inner Mbugu It is not my intention to claim that the source language was (Borana) Oromo. Indeed, there is evidence that it was not. For example, the Inner Mbugu word ka-wahá 'knife' is from the "other" Cushitic source, from proto-Eastern Cushitic *warhhan 'spear' — warhhan 'knife' in Rendille — but not from Oromo since Oromo has woraana (Sasse 1979:37). There is no evidence for a Southern Cushitic source for this root. A complicating factor is that the exact position of Southern Cushitic within Cushitic is also disputed. Greenberg had posited separate branches, East, South, in addition to Central, North and West, which in my opinion boils down to stating that all these groups are Cushitic but that their subclassification is uncertain. Hetzron (1980) has argued on morphological grounds that Southern Cushitic is part of Eastern Cushitic. It will be clear that it is impossible to make steadfast statements at this stage about the exact position within Cushitic of this Kenyan Cushitic link of Inner Mbugu. Therefore I use the term Old Kenyan Cushitic, which is neutral in a classificatory sense, for any Cushitic evidence that is clearly not related to the modern Southern Cushitic languages (esp. Gorwaa/Iraqw) and that can be shown by comparison with Eastern Cushitic languages. Representatives of Old Kenyan Cushitic are Dahalo, Yaaku, Taita Cushitic (see 2.7), and possibly proto-Baz. The latter two, Taita Cushitic and proto-Baz, are postulated on the basis of loanword evidence by Ehret and Nurse (1981) and Heine et al (1979) respectively. It would be premature to decide whether the evidence from Oromo and Dahalo, and of Taita Cushitic, reflects one single Cushitic source of Inner Mbugu. Dahalo is another good candidate to prove a second Cushitic link with Inner Mbugu, on both the lexical (cf. Table 2.11) and grammatical levels. Dahalo has been classified as the first branch of Southern Cushitic by Greenberg (1966) and Ehret (1980) but this classification has been challenged by Tosco who has done the most recent and extensive research on this language (see Tosco 1989, 1991, 1992). The al­ ternative that he proposes is a closer link between Dahalo and Yaaku/Mokogodo, which is classified as Eastern Cushitic. The Dahalo now live on Lamu island in Kenya and on the coast opposite the island. It has been argued, however, that these people once spoke a Khoi-San language and that the shift to a Cushitic language took place in Central Kenya (see Nurse 1986).

HISTORICAL AND GEOGRAPHICAL BACKGROUND

39

Dahalo link Inner Mbugu ' óro (v) 'wear, close' Tur 'sew' ára 'white-hair' 'áára; cf. Oromo (h)arrii, and Sagala lu-'ara bále (v) 'move house' b 'alak báyu (aj) 'red' b'aba 'to shine' bó (v) 'rot' p 'Ó' ate bódi (v) 'sleep' b'- 'sleep' + di 'down' búku (v) 'draw (water)' b'ukul 'fill a hole' búsu (v) 'curse' b'ut'uïl spit' bútu (v) 'put into', bu(')u 'to enter' bu' 'put into', bu'ut 'enter' dama (v) 'collect fruit' d'ar- 'put together, collect' girú (aj) 'big' gúrume 'old (of persons)' gomaè 'clothing' góóra 'piece of cloth' gú (v) 'uncover, come out' gud- 'remove' gu'á 'pig' guáát'e (Ehret 1974) hé (v) 'arrive' ha? 'come, arrive' hlahá (v) 'bum (tr)' hlahh- 'put on top of fire' 19 hlámu (v) 'sweet' amalgamation of Dahalo hlááhhame and Swahili tamu igwahlú also i' ohlu 'cheek' ngoohlo 'soft palate' káka (v) 'divide' k'akk'aatad- 'divide' (frequentative) kéla 'hunger' A'ére 'hunger' kiharige 'tail' rik'a 'tail' kirii (v) 'return' kiir- 'go and come back' kola (v) 'meet' Dahalo or Somali kulum korra 'public meeting' kuhamé (aj) 'black' himmate plus ku- adjectival prefix m-'á 'head' Tani 'head' m-buratú 'boy' b'óór-eete 'boy' and -to singulative suffix ma'ilíma 'tears' 'ilima; cf. Oromo ilm-aani, Iraqw ilmo 'drop' makahlú 'dream' milaak'-ani, with metathesis mhlaé or mhlayé 'thorn' mágohle, with metathesis wkorá 'doctor, circumcised person' £'oro 'tree' (tree-medicine) na'á 'honey' nala 'honey' på (v) 'eat' po'eem 'be consumed' ro - ' g o ' ró (v) 'leave behind' sahhan- 'greet' sähe (v) 'meet' t'a?- 'to poison' tå (v) 'cheat' tarar- 'tremble', tar- 'mix' torá (v) 'shake' Table 2.11: Inner Mbugu words linked to Dahalo The Yaaku, who live in the Laikipia area, which is the' legendary homeland of the Mbugu, have now shifted to Maasai (Brenzinger 1992). The remnants of their lan-

40

CHAPTER 2

guage have been described by Heine (1975). The Yaaku were a group of huntergatherers who have recently taken to herding (Cronk 1989). According to Greenberg's (1963) classification, their language is Eastern Cushitic. As was already observed by Feierman (1974:75), there are a number of lexical items in Inner Mbugu that show a remarkable resemblance to Yaaku lexical items. Some of the proposals he makes, however, have alternative and better etymologies. These and others are given in Table 2.12. 'ihlé (aj) 'unripe, bad'; cf. Yaaku 'ihle' (m) 'grass sp., pasture'; alternatively from Iraqw eetl 'fresh' (not reconstructable for West-Rift Southern Cushitic). gú (v) 'uncover, come out' cf. Yaaku gu' 'uproot' or Dahalo gud- 'remove'. i-di 'land' most likely from Iraqw di 'place' but Yaaku dimin 'place' is an alternative. kakéra 'side' and kará 'front, sideways'; cf. Yaaku kaar ' 'side' or Oromo  'arree 'edge'. ké (v) 'put'; cf. Yaaku ké?s (v) 'put up, plant', or Oromo kaa, keeta 'put something somewhere', or Dahalo kaaj-. kó (v) 'hear'; cf. Yaaku ókó 'voice, word'. kulelé 'kid of goat or sheep'; cf. Yaaku kollεh 'castrated goat', Oromo lalleesa 'billy goat'. Ió (v) 'have'; cf. Yaaku lo ' ' 'have'. nuŋa 'nose'; cf. Yaaku ú possibly conflated with Iraqw dunga' (West-Rift urung-). tú (v) 'pound, dig'; cf. Yaaku tuu' 'pound', Oromo tuma 'beat', Iraqw tu'-uut. rú (v) 'be ill' and u-rukáo 'illness'; cf. Dahalo ruk' o 'illness', Yaaku lu' 'be ill', Oromo d'ukkuba 'illness', Ilwana roghoti 'illness'. wá 'cow(s)'; cf. Yaaku wàa'; alternatively Iraqw/Gorwaa hikwa / yakwa 'cow(s)'. wé 'one'; cf. Yaaku wehet 'one', or NMb mwe Table 2.12: Yaaku - Inner Mbugu lexical resemblances In addition, two other possible links with Yaaku can be mentioned: One of the surpris­ ing derivations in Inner Mbugu is that the verb 'to close' is derived from the verb 'to open' by a causative; the Yaaku verb 'to close', deps, has a final s which looks like a frozen causative. Secondly, and very speculatively, the addition of 'u to the demon­ stratives in the Bumbuli dialect of Inner Mbugu might be related to the final elements a'a, and in particular a 'au in plural demonstratives in Yaaku (Heine 1975:41). The Yaaku were until recently a kind of servant group of hunter-gatherers (Dorobo), a fact which does not fit with the culture of the Mbugu. However, there are several ways in which this discrepancy can be explained. One possibility is that the Yaaku were an impoverished people who used to have cattle in addition to agriculture. Another option is that there was a Cushitic speaking group (the forefathers of the Mbugu) who had the Yaaku as servants and that the Yaaku took over the language of their masters, as they have done recently again by shifting to Maasai. I take the Yaaku evidence as pointing to an Old Kenyan Cushitic link, rather than specifically Yaaku.

HISTORICAL AND GEOGRAPHICAL BACKGROUND

41

If there once was a Cushitic language pre-Ma'á, this language was most proba­ bly part of the Old Kenyan Cushitic presence in Kenya described above and would best be called Old Kenyan Cushitic until the subclassification of Cushitic is resolved. The Inner Mbugu items that I have indicated as having an Eastern Cushitic source in Inner Mbugu —that is, Oromo, Dahalo, Yaaku and also some Sagala and Davidąare in that case to be understood as retentions of the Cushitic pre-Ma'á. We should not forget that there is evidence that the Dahalo once lived in Central Kenya and that a link between Yaaku and Dahalo has been proposed. It is quite possible that the former Taita Cushites are also part of this complex. We have seen in sections 2.3 - 2.8 that the lexical evidence supports the oral traditions in the history of migrations with evidence for a link to the Pare area, the Maasai, and the Mbugwe/Gorwaa area. In addition, the lexical evidence suggests a link to Old Kenyan Cushitic. In the next section we look at the lexical material again, this time to establish the origin of the Inner Mbugu words in certain domains, in an attempt to find evidence for the chronological order of the links to these various areas. 2.9

Chronology through a study of lexical domains In this section I investigate the origin of the Ma'á (Inner Mbugu) vocabulary in several lexical subdomains in order to find clues for the chronology of influence from the several areas that were discussed in sections 2.3 - 2.8. For the names of the major neighbouring peoples there is a separate Inner Mbugu term. The Shambaa are called m-sitá (class 1/2) in Inner Mbugu. The origin of this word is unclear but Meinhof had va-sitane. The ending ne appears as ni in Maasai on the singular of ethnic names, e.g., ol-máásani 'a (male) Maasai' (also ol-maasinta) and reappears in the ending nthi of the form ki-maanthi 'Ma'á language' in Meinhof (1906). The autonym m-Ma'á (1/2) 'Mbugu' is most likely from the Maasai autonym Maa 'Maasai'; variation between long vowel and two identical vowels with an inter­ vening glottal stop is common in some Cushitic languages, e.g., in Oromo.20 With par­ allel lexicon building in mind this evidence (the ending of these two words and the authonym Ma' 'a) suggests Maasai influence at the time when the need for a word for the Shambaa and for a different autonym was felt; that is to say, the parallel lexicon probably developed after having left Maasai domination and during the period of set­ tlement in the Usambara mountains. The word for Pare is m-'ariyé (1/2) and this is possibly related to Inner Mbugu m-'áro (1/2) 'person who is similar or comparable (to you)' but the origin of this last word remains unclear. The Normal Mbugu terms are all from Pare: m-shamba (1/2) Shambaa from Pare m-shamba and not from Shambaa mu-shambala; mw-ásu (1/2) 'Pare' is also Pare. The words for the clans of the Mbugu people are not double in Inner and Normal Mbugu presumably because the use of them already signals Mbugu ethnicity. Names in general are common to both Inner and Normal Mbugu, and show the usual local Bantu structure with the proclitics se and na (see section 8.16.7). There are no personal names that point to a non-Bantu origin. Words denoting family relations, all in the class pair la/2 (see section 6.1) are parallel, but the lexicalization patterns follow in detail that of Pare.

42

CHAPTER 2

Most of the wild animals, insects, birds and uncultivated plants are identical in Inner and Normal Mbugu. Quite often they are not from Pare origin but from Shambaa or Swahili, signs of recent borrowing. This is probably an effect of the multitude of languages mastered by the speech community: Inner Mbugu, Normal Mbugu, Sham­ baa, Swahili and Pare. It will be clear that given the fact that these Bantu languages are closely related it is not always possible to know whether the origin is Swahili, Shambaa or Pare; in the etymologies, Shambaa or Swahili are only given as the source in the following tables when Pare does not fit. The conclusion is that the words in the more specialised areas of the lexicon tend to come from the dominant language in the area: Shambaa. A second conclusion is that the Mbugu were most likely not a group of hunters in their recent history; otherwise we would expect them to have had Inner Mbugu words for these common and prominent animals. In Table 2.13, Sw refers to Swahili, Pa to Pare, Sh to Shambaa and I+NMb to words that are common to Inner and Normal Mbugu (NMb). Mbugu nhtwiga 'giraffe' mbogho 'buffalo' nhkímá 'monkey' nhkimé 'monkey' nhkoo 'vervet monkey' nhtumbiri 'vervet monkey' mbegha 'colobus monkey' nhkendá 'zorilla?' nhkánga 'turkey' nhkálá 'crab' ngwirizi 'eagle'

language I+NMb I+NMb NMb 1Mb I+NMb I+NMb I+NMb I+NMb I+NMb I+NMb I+NMb

source Sw: twiga, cf. Pa: hori Sw: mbogo or Pa: mboghó Pa: nkima NMb: nhkima Pa: nkó Pa: ntumbiri, Sw: tumbili Pa: mbegha Sh: nkenda 'squirrel' Pa: nkànga Pa: nkàla Pa: ngwirini

Table 2.13: Sample of fauna vocabulary and its origin Banana culture is something new and foreign to the Mbugu. This could already have been deduced from the short text in section 1.6 which relates how the Mbugu planted the banana trees upside-down. The banana terminology is relatively limited and mostly not doubled into two parallel forms. In Table 2.14 only the first five items have an equivalent in Inner Mbugu. The sources for the (Inner+)Normal Mbugu terms are partly from Shambaa, a fact which suggests that they were acquired recently (Sh= Shambaa, Pa = Pare). NMb idafá idiyá ij im ikinda

meaning dried banana bark ripe banana banana leaf banana tree

1Mb ilamba igeru ireno ikindeno

source for NMb Pa: idava Pa: idió 'banane' Pa: ijitu. S.Pa: nhkínda

HISTORICAL AND GEOGRAPHICAL BACKGROUND

NMb itambwe itonhté ikiri ishumba ishwagho kigógo lukamba mughunda

meaning ensete, false banana banana banana tree after harvest banana flower dried banana leaf bare stalk of banana bananas in a bunch field of bananas

1Mb m-viro

-

43

source for NMb Sh: tambwe Sh/Pa tonte, ma- 'ripe dessert banana' 21 Sh: kili Sh: shumba 'male bud of the Banana' Sh: shwagho Pa: igogo. Pa/Sh: lukamba Sh: m(u)ghunda 'banana plantation' 22

Table 2.14: Mbugu banana terminology with sources for the NMb terms Most intriguing are those five banana words that do have a special Inner Mbugu form. Ilámba or ilambeno is a recent innovation from Shambaa lamba 'dried banana bark', due to the fact that bananas and banana tree were obtained from Shambaa; it is parallel to the NMb form from Pare. It has a variant with the ending -éno, just like the word for 'banana tree', ikindéno, which is derived from the Normal Mbugu/Pare ikinda (see section 3.7). Both facts support the claim that banana terminology is recent in Mbugu but also that the double lexicon was operative at the time of the introduction of ba­ nanas, given the Inner Mbugu equivalents. The most deviant forms are those for very common concepts such as 'ripe banana' and 'banana leaf: igerú 'banana (ripe)' from Gweno i-rigo 'wild banana plant, ensete' with metathesis; iréno 'banana leaf from Gweno iru 'banana leaf plus the ending -eno. Both suggest an origin that is Gweno in North Pare. There is also m-viro 'false banana, ensete', possibly from Saghala viro 'maize grains or grains of other cereals'.23 Cattle terminology is much more developed than banana terminology and shows an Inner Mbugu lexicon that is never derived from Normal Mbugu and which is seldom shared, i.e. even for the most detailed terms there is a double lexicon; or, ex­ ceptionally, there are some distinctions that are only made in the Inner Mbugu lexicon. For example Inner Mbugu has a separate word for a castrated goat, ogu, and a heifer, mili, two words that have no equivalents in Normal Mbugu. Where Normal Mbugu has a derived noun, kabuji for a 'small goat', Inner Mbugu has an underived, different lexeme, kulele, which seems to be from the same source as kulelo 'young female goat'. Both the cultural information and the structural properties of the lexical parallellism suggest that these words potentially represent the oldest lexical layer. The origin of the more detailed cattle terminology in Inner Mbugu (Ma'á) is mostly obscure; some of it can tentatively be linked to Old Kenyan Cushitic. This would fit with the oral traditions if we consider the cattle culture as given and would make the Old Ken­ yan Cushitic layer the oldest lexical layer. i-'alu (5/6) 'sheep'. From a source close to Oromo (Borana) (h)oolaa with metathesis of the vowels. The NMb equivalent igónji (5/6) 'sheep' from Pare igónji 'sheep' is ultimately Cushitic, as in Southern Cushitic West-Rift gwand-.

44

CHAPTER 2

alé (9/6) 'ram'. Possibly from the same source as i'alú with addition of a éCV suffix and truncation (see 3.4); or, alternatively, from Iraqw lee 'i 'goat' (WR: lege'i) The NMb equivalent idóromé (5/6) 'ram' is from Pare idóromé 'ram'. m-'úse (3/4) 'sheep that has not calved yet' From Davida osi 'ewe, lamb'. The NMb equivalent is muhati (3/4). afá (9/10) 'goat'. Origin unclear. The NMb equivalent is mbuji (9/10) from Pare mbúji. dé (9(/10)) in Bumbuli 'cow' sg, in Magamba 'goat(s)'. The singular/plural pair dé / wa in Bumbuli is a reflex of Iraqw (or West-Rift Southern Cushitic) hlee I hikwa; the meaning 'goat(s)' in Magamba is a reflex of Oromo ree or dee 'goats'. hlané (9/10) 'male goat'. Origin unclear. The NMb equivalent is ndenge (9/10). ogú (9/10) 'male, castrated goat'. Origin is unclear (Iraqw gurta, Maasai ol-oro). There is no NMb equivalent. amo 'goat'. Occurs in Meinhof (1906), Green (1958). In disuse. kulelé (9/10) (also 12/13) 'kid of goat or sheep'. Possibly from Oromo lalleesa 'billy goat'. The NMb equivalent is katangina (12/13) (also 9/10) 'kid of goat, sheep; child' from Pare katangina 'kid of goat' or else kabúji (12/13) 'small goat' is used which is derived from mbuji 'goat'. kulélo (9/10) (also 12/13) 'young female goat'. Cf. kūlele. The NMb equivalent is mbáríka (9/10) (also 12/13) 'young female goat' from Pare mbáríká 'young female goat*. mbúru (9/10) 'tail of a goat'. Origin unclear. In NMb kishambé (7/8) also kishambi, is used, from Pare mshambi 'tail of a sheep'; this NMb word is sometimes also used in 1Mb, sometimes as mshambeno. idehedehe (5/6) 'dewlap (of sheep, goat, cow)'. Origin unclear. No NMb eqivalent is known. iburú (5/6) 'droppings of goats, sheep'. Plural is more common. Ehret links it with Dahalo búrune 'dust'. No NMb eqivalent is known. ahla'u (9/10) 'room for goats'. From 1Mb ahla 'fire' (goats are kept in the area of the house which is close to the fire); the NMb equivalent, nyumbeni, is the word for 'house' with a locative clitic. Table 2.15: Ma 'a (Inner Mbugu) goat/sheep vocabulary and its origin 2.10 Reconstructing the history: possible scenarios In trying to understand how a mixed language such as Ma'á came about one would like to link the linguistic history with the history of the people. In the following I will discuss the historical knowledge of the area (and the lack thereof). After discuss­ ing some background conditions I will present a number of crucial facts that are to be explained by the history of the Mbugu (language and speakers) and finally I will raise the two crucial issues: Where and when did the mixed variety develop and did the Mbugu once speak a Cushitic language, and if so, where and which one?

HISTORICAL AND GEOGRAPHICAL BACKGROUND

45

The history of the Mbugu people is linked to the history of the surrounding people and in particular to that of the Shambaa, the Pare, the Taita, the Mbugwe, the Maasai, and the Dorobo groups.24 For all of them there are no historical sources for the period that is crucial for the development of Ma'á/Mbugu, i.e. the period previous to and including the first half of the 19th century. Archaeological evidence in the rele­ vant areas is scarce and concerns periods much earlier than that. The only sources of information next to language are oral histories and analyses of cultural elements. For­ tunately, there are a number of excellent historical studies of most of these groups: Winans (1962), Feierman (1974), Conte (1994), and Huijzenveld (1997) for Shambaa; Kimambo (1969) for Pare; Harris (1978) for Taita; Spear and Waller (1993) and refer­ ences therein for Maasai; and Walsh (1990, 1992/93) for the Dorobo of the area. One of the fundamental questions underlying a discussion of the history of Ma'á language and people is: To what extent can we distil former social situations from historical linguistics? It is unrealistic to strive at predictability of former social situations from historical linguistic insights; still there are clearly certain social situa­ tions that favour certain linguistic outcomes. Bakker (1997) has proposed two kinds of socio-linguistic situations in which mixed languages typically arise. One is a society in which the mothers are from one language group and the fathers are from another, suf­ ficiently different language group. The second sociolinguistic situation which favours the development of a mixed language is that of an itinerant occupational group. It does not follow that Ma'á should by necessity go back to one of these sociolinguistic situa­ tions but independent evidence for such a historical situation would be strengthened by Bakker's observation on the origin of mixed languages. Against this background, possible scenarios for the history of the Mbugu people can be developed. In comparing possible scenarios we try to minimise the number of putative movements and to maximise the number of ingredients that need explanation. In the following list of the various ingredients that the ideal scenario should include 111 are linguistic, 12-15 are historical, 16-20 are cultural and/or socio-economic: 1. The people felt a need to build a full parallel lexicon in order to mark an identity that is markedly distinct from that of neighbouring Bantu agriculturalists. 2. Inner Mbugu contains elements from Gorwaa/Iraqw (see 2.5). We can deduce from this that the Mbugu people or a proportion of them were at one point in con­ tact with Gorwaa/Iraqw; this would be the area of the present-day Mbugwe people (cf. ingredient 10). Alternatively or additionally a number of Gorwaa entered the Mbugu population (cf. ingredient 3). 3. One of the significant words from Gorwaa/Iraqw is m-lagé which means 'mother' in Inner Mbugu and 'cow acquired in war' in Iraqw (and presumably Gorwaa). This suggests that a number of women entered Mbugu society at times of conflict between the Mbugu and the Gorwaa/Iraqw. Even a few decades ago, a Gorwaa family moved to the Mbugu in the Usambaras (Nordbustad p.c.). This ingredient depends of course on the question whether one accepts the proposed etymology.

46

CHAPTER 2

4. The Ma'á parallel lexicon has a considerable amount of Maasai vocabulary (see 2.4). The important role of Maasai is further underlined by the fact that a rela­ tively high proportion of the Maasai lexical elements are verbs which is atypical for normal additive borrowing. Some of the semantically empty nominal suffixes could have their origin in a reinterpretation of Maasai endings (see sections 3.73.8). At least one of the lexical items from Maasai, the word for 'dog', is shared by Mbugwe (see section 2.5). The linguistic influence of Maasai correlates with ingredient 11. 5. The parallel Ma'á lexicon has Cushitic elements from Old Kenyan Cushitic (see section 2.8). 6. There are a number of elements in the Ma'á lexicon that are also found in Sagala (Taita Bantu) (see section 2.7). The Ma'á lateral fricative can be linked to Sagala. The links with the Taita hills can be relatively recent or very old. There is mention of a migration to the Pare mountains from the Taita area and, in farily recent times, to the Usambara mountains. 7. The Pare elements in the Normal Mbugu lexicon and in the common grammar do not come from one single Pare dialect; there are clearly elements from North Pare in addition to the majority from South Pare (see section 2.6). The North Pare con­ nection is further strengthened by some similarities with Gweno, which is and was only spoken in North Pare (see section 2.6). 8. Dialect variation in Mbugu is limited to the Inner Mbugu lexicon and the dialect distribution suggests stronger Iraqw/Gorwaa presence in the Bumbuli area. 9. The degree of knowledge of Ma'á is not uniform across the Mbugu speaking area. It is strongest in Bumbuli and in the western part of the larger Magamba area, and dependent on clan affiliation in Rangwi. These facts correlate with the route of en­ try, see Table 2.1. 10. The name Mbugu is identical to the name Mbugwe, a Bantu group (F.34) living North of Babati.25 The similarity in name could be accidental and the basis for oral traditions that link the Mbugu with the Mbugwe. If it is not accidental, the similar­ ity in name correlates with the oral history. 11. The name Ma'á is similar to Maa, the name of the Maasai language. 12. There are two main versions of the most recent part of the orally transmitted his­ tory of the Mbugu. One is a movement from South Pare to Usambara. This in fact corresponds to a general movement of population of the Pare people as well. The other version is one of leaving the Pare mountains in pursuit of cattle stolen by the Maasai, a period of serfdom among the Maasai, robbing Maasai cows and driving them through Handeni to the Usambara mountains, this time with the Maasai in pursuit. 13. The oral histories mention their ultimate origin as being in a place called Laikipia (=Lukipia) in Kenya. This is the place from where the Tanzanian Maasai departed (see section 2.1) and where the Yaaku people still live. Statements in the first lines of oral histories about the place of ultimate origin are known to be particularly un­ reliable; the Mbugu could simply have copied Maasai origin myths. However, Ya-

HISTORICAL AND GEOGRAPHICAL BACKGROUND

14.

15.

16. 17. 18.

47

aku is a possible source of the Old Kenyan Cushitic elements in Ma'á, cf. ingredi­ ent 5. The entry of Mbugu into Usambara is both through Rangwi and through Handeni. Thus there were at least two different groups before entry into the Usambaras. Ei­ ther they first became one new ethnic unit Mbugu in the Usambaras or they found each other again. According to some oral traditions, the Mbugu were once a client group of the Maasai. Various such groups exist and they are usually called Dorobo. This could link the Mbugu to the Southern Cushitic Aasáx who were/are such a Maasai client group; this could explain some of the Cushitic elements in Ma'á. Otherwise, this could offer possibilities of linguistic contact with a variety of sources and un­ common linguistic behaviour as such groups often are of mixed ethnic back­ ground. The Mbugu people have somatic features that are to some extent recognisably dif­ ferent from their neighbours. This strengthens the assumption that at least some of their forefathers once spoke a non-Bantu language. The economic activity that is central to Mbugu culture is cattle keeping. Initiation camps were (are?) in two areas. The fact that there were two different places correlates with ingredient 14. Initiation was relatively short. There is no in­ dication that a secret language is learnt during initiation. In general, a secret lan­ guage acquired during initiation takes the form of a parallel lexicon. In principle, the origin of a parallel lexicon could lie in such an initiation language, provided that the initiaton period used to be longer. The Pare had at one point a longer ini­ tiation in which they learned a secret language.

Keeping the above ingredients in mind we can now start to build some possible sce­ narios of Ma'á formation starting with two central questions in the history of the Ma'á language and people: Where and when did the parallel lexicon Ma'á develop? and Did the Mbugu people ever speak a Cushitic language and if so which one, where and when? First scenario: The parallel lexicon developed in the Usambaras. Two groups came together, one speaking Mbugu-Pare, the other a heterogeneous group speaking Maasai, Gorwaa, and possibly Aasáx. These came to form one people because of the cultural similarity between the two groups, because both were minorities and both were resistant to farming. Thomason (1995) distinguishes a dichotomy of mixed lan­ guages: those that originate gradually as opposed to those that originate in a short pe­ riod of time. The latter type of mixed language is typically created by a new ethnic group. Especially for this group of mixed languages she recognises creative linguistic behaviour outside the usual domain of language change as a factor in its emergence. She mentions language games of the 'secret code' variety. Thomason places Inner Mbugu/Ma'á in her first group and does not consider such paths for the emergence of Ma'á. However, the structure of the double Inner Mbugu vs. Normal Mbugu lexicon shows that "creative linguistic behaviour" by way of register creation did play an im-

48

CHAPTER 2

portant role in Inner Mbugu/Ma'á too. Moreover, there is strong evidence from the oral traditions and the geographical distribution of clans and language behaviour that the Mbugu entered the Usambara mountains in at least two different population moves and fused to form one ethnic community. If these two groups were different in origin, then the extra lexicon must have arisen in the Usambara mountains and the oral tradi­ tions were adjusted to stress a unity of the newly-formed ethnic entity. This scenario explains why there is no trace of a Mbugu language in the Pare area. This does not directly explain how the language acquired the Old Kenyan Cushitic elements. A more likely option is that the parallel lexicon developed shortly after their stay among the Maasai. A large portion of the group were already Mbugu/Pare speak­ ers. Other members from the group came from Gorwaa and the two groups fused as a servant group of the Maasai. This scenario would explain the Maasai influence. The Old Kenyan Cushitic elements could be remembered words from a former language of those who at that time already spoke Mbugu-Pare. This scenario stays close to the oral traditions, namely that both groups ultimately came from the Pare mountains and that they had already been one ethnic entity there; it also explains better the similarity be­ tween the name Mbugu (also present in the Pare mountains) and the name Mbugwe. But if the two groups were ultimately (at least partly) related, the parallel lexicon can have been formed at any point in place and time between the Mbugwe/Gorwaa area and the Usambaras. Ideally, a closer examination of Mbugwe would reveal such a link between Mbugwe and Pare. The emergence of the parallel lexicon is then not only related to ethnicity but also to efforts to communicate. In the Usambaras it acquired its status as an ethnic emblem. Possibly, the parallel lexicon emerged out of mixed mar­ riages between Gorwaa men and Mbugu women, either in the Mbugwe area or in the Maasai plains. The children learned Mbugu but not enough Gorwaa. Such a scenario of a community consisting of mixed marriages is known to give rise to a mixed lan­ guage, see Bakker (1997). It is consistent with ingredient 3. The parallel lexicon need not have been created from scratch. A situation of language shift may have laid the foundation of a parallel vocabulary. The forefathers may have shifted to the Bantu language while remembering some, basically core, cul­ tural lexical elements as well as pronouns from the older language which they restruc­ tured to the Bantu (Pare) frame. Later this limited choice was deliberately enlarged in order to create their own "language". The Mbugu shifted to Pare, keeping their rudi­ mentary Cushitic language as an in-group code. Possibly their former Cushitic lan­ guage survived in an initiation language which in the Bantu environment was used an in-group code. If such an initiation language existed it could have provided the cir­ cumstances for lexical expansion. Due to a period in a non-Bantu environment (Maasai) and the influx from Cushitic (Gorwaa) speakers they were able to expand that code drastically, keeping it different from Mbugu-Pare. The other question is: Did the Mbugu ever speak a Cushitic language? And if so, where and when? Taking the first question we can state that the number of Gorwaa elements is high enough to assume that at least some of the Mbugu had knowledge of this language, but possibly only as a second language. The non-Gorwaa Cushitic in-

HISTORICAL AND GEOGRAPHICAL BACKGROUND

49

fluence on pronouns and the //-causative suggest that at least some of them once knew a different Cushitic language. Some of the lexical data in Ma'á from Old Kenyan Cushitic are pre-Oromo. Since there is no indication for recent important contacts be­ tween Mbugu and present-day Kenyan Cushitic languages such as Yaaku or Dahalo, this must go back to an earlier period. The Pare oral traditions and the Mbugu oral tra­ ditions agree on the fact that the Mbugu already found Pare/Gweno when they arrived in the Pare mountains. If this interpretation of the oral history is correct, then it is logi­ cal that they spoke this Cushitic language before arrival in the Pare mountains, as the shift to Pare must have occurerred after their arrival. There are several candidates for the place where this language could have been spoken. Most positive evidence is pre­ sent for area (b) below but this does not exclude areas (a) nor (c): a) Laikipia in accordance with the Mbugu myths of origin. In Laikipia there are still remains of a Cushitic language, namely Yaaku. b) The Taita hills. There is abundant evidence for a former Cushitic population in the Taita hills and there is a link with the Pare area since, according to the Pare tradi­ tions, some the Pare people have their origin in the Taita hills. c) Somewhere in the Maasai plains. Here the Mbugu would have been a Cushitic speaking Dorobo group such as the Aasáx. However, they would have had cattle since, according to the oral traditions, the Mbugu already had cattle when they en­ tered the Pare mountains, apparently more than the resident Bantu speakers. Concluding, a scenario that combines most of the considerations above is the follow­ ing: In the Pare mountains there was once an (Old Kenyan) Cushitic speaking group (linked to a similar group in the Taita hills) who shifted to Pare (Chasu). Part of this Cushitic population left the Pare mountains for the Maasai plains. Those who re­ mained in the Pare mountains completed the shift fully. Some of these, the ones who completed the shift, later went straight to the Usambara mountains. They consider themselves Mbugu and are considered by the other Mbugu as such. At least the expan­ sion of the parallel lexicon took place shortly after (some of the) Mbugu formed a ser­ vant group among the Maasai. This group had a considerable influx of Gorwaa people. Possibly this influx happened in an earlier period when they were living in the Mbugwe area or, otherwise, the influx (of Gorwaa men) occurred in the Maasai plains. In the Usambara mountains, the two groups (the ones who came through the Maasai plains and those who came straight from the Pare mountains) met and reconstituted one single ethnic group. From this time the Normal Mbugu (Pare) language had an influence on lexical restructuring. In the Maasai-dominated group the Normal Mbugu (Pare)-speaking women were in the majority, determining the grammar of the lan­ guage (see Winter, 1979, for the sociolinguistics of a similar servant group). Nobody in the servant group had loyalty to Normal Mbugu/Pare, although there were people who were bilingual in Normal Mbugu/Pare. Emotionally, however, they were attached to their former language which they considered to be linked to cattle culture. Cattle culture was the norm in a Maasai dominated environment. The use of non-Pare lexis was favoured by both the people (fathers) of the Gorwaa influx, and the people who

50

CHAPTER 2

came from the Pare area. Possibly an initiation language in Vudee (South Pare) where the "Maasai"-Mbugu came together was influential in the building of the parallel lexi­ con

3

LINGUISTIC HISTORY

3.1

The issues and the debate Given the lack of early written sources, the synchronic situation is the point of departure for research on the linguistic history. As a consequence it is important to look into the developments of Ma'á and Mbugu during the period for which we do have material. In particular there is the logical possibility that the present-day situation of Ma'á being a parallel lexicon to Normal Mbugu is a recent development. In section 3.2 I argue that this is most likely not the case and in sections 3.3 and 3.4 I look into the historical processes that the documented history does reveal. A number of issues in the linguistic history of Ma'á are controversial and these will be discussed in this chapter. The first issue is that of the nature of the historical developments and in par­ ticular the role of conscious lexical manipulation in the emergence of Ma'á; a second issue is the question of the need to postulate a language shift. Another option is grad­ ual borrowing of Bantu grammar as argued by Thomason. In order to address this is­ sue, the possible remnants of non-Bantu grammar in Ma'á are discussed in detail in sections 3.5-3.9. In section 3.101 argue for the need to postulate a shift from a former Cushtic language. A final issue is that of the role played by code-switching in the development of Ma'á. Myers-Scotton has proposed that Ma'á is the result of an arrested turn-over in matrix language. In section 3.111 argue that her code-switching model is insightful for an understanding of the mixture that Ma'á speech presents but that there is no evidence for a decisive role of code-switching as such in the development of Ma'á structures. Finally, in the last section of this chapter (section 3.12) I discuss what we can learn about historical linguistics from the Ma'á case. 3.2

Absence of drastic recent linguistic developments Thomason (1997) has raised the question whether the linguistic situation has drastically changed since the late nineteenth century. She hints at this possibility to explain the fact that the earlier accounts of the language do not mention the double nature of the lexicon nor the existence of two languages/registers. My explanation is that the earlier researchers concentrated on the most interesting topic, i.e. on the devi­ ant language, and did not consider it important to ask about or to report on the ordi­ nary Bantu language. Thus, both Baumann (1890) and Copland (1933/34) collected tales, songs, and proverbs from the Mbugu and found out that all the material turned out to be Pare (Brenzinger 1987:165, Copland 1933/34:243), or more probably Nor­ mal Mbugu and they refrained from publishing it. Copland managed to get one Ma'á text, on the origin of the Mbugu, and that is the only text that he published (Copland 1933/34). The text is pure Inner Mbugu and is not different from present-day Ma'á. Going back to the earliest sources, we find that the words in Farler (1885) and Storch (1895) are all Ma'á. From this we can only conclude that the speakers mastered Ma'á and that the Ma'á ofthat period was not very different from the present-day language, at least not in the core vocabulary that they published. In the same period Baumann

52

CHAPTER3

(1890) found Mbugu who spoke Pare (presumably our Normal Mbugu). Either there were differences among the ethnic Mbugu, some speaking Normal Mbugu and others speaking Inner Mbugu (Ma'á), or the situation was basically identical to what it is to­ day: the people speak both varieties, consider the Inner Mbugu to be the language that is really theirs but rely on Normal Mbugu for folktales and other verbal art. The next scholar on the scene is Meinhof who spent some time at the mission station of the missionary Roeh in Bumbuli and travelled around in the area. It is not clear how much time he spent working on Mbugu. From the recordings of various languages that he made on wax rolls, we can deduce that he spent at least a month from mid-November till mid-December 1902 in Bumbuli. He recorded a Mbugu song and a text in Tanga in February 1903.' The language that Meinhof describes is Ma'á (Inner Mbugu), or Ma'ánthi as he calls it. This latter denomination is still in use in the Bumbuli area. He does not make mention of the "normal" Mbugu language but he does make the remark that it is possible that the Mbugu use more Bantu forms when speaking to Bantu people. This remark is made, however, in connection with the varia­ tion of words with or without noun prefixes (see section 3.5) and not with reference to different registers or different languages. The "normal" Mbugu speech must have ex­ isted at that time already because Meinhof did collect some Normal Mbugu items without recognizing them as such. For example, he does have the Normal Mbugu word vijú 'food' instead of the Inner Mbugu word vi'aghú and he got this in a com­ pletely NMb sentence, alavidju 'sie ißt' [= ala vijú 's/he eats'] and nalavidju 'ich esse Speise' [= n-a-la vijú 'I eat food'], not as a separate word (Meinhof 1906:307). Other examples are khingo 'Fell auf dem man schläft' [nhkíngo 'hide, skin'] instead of 1Mb buhlé, and m'kota 'unbearbeitetes Fell' = mkota (3/4) 'skin (not-dried)' [NMb], in­ stead of mkwano [1Mb]. Meinhof has the Normal Mbugu word nyoka 'Schlange' for 1Mb borné 'snake'. The 1Mb word borné 'snake' was already reported by Storch in 1895, a source that Meinhof did not use. It is thus most likely that both Inner Mbugu and Normal Mbugu were already present at Meinhofs time. In the case of the word for 'rain', he has both forms: mare [1Mb] and mbua [NMb]. The grammar and lexicon in Meinhofs article are that of present-day Ma'á of Bumbuli, that is, the productive grammar is identical to that of Normal Mbugu (and Pare) and the lexicon is different from Normal Mbugu and more or less identical to present-day Inner Mbugu. Most likely the speakers also spoke Normal Mbugu, which would explain the intrusion of some Normal Mbugu words in his material, and presumably the two lexica were grosso modo parallel to each other as they are now. 3.3

History of the lateral fricatives in Ma'á (Inner Mbugu) The lateral fricative is without doubt the most characteristic sound in Inner Mbugu. The sound is not common in East Africa2 and it is considered to be character­ istic for Southern Cushitic, and for Dahalo. The Southern Cushitic languages have a (ejective) lateral affricate in addition to a voiceless lateral fricative.3 In a limited num­ ber of cases the Inner Mbugu word is derived from the Normal Mbugu counterpart by replacing a sound with this lateral fricative. Therefore I will study the origin of this

53

LINGUISTIC HISTORY

sound in Ma'á and its development through history in some detail. We will see on the one hand that at one point in history Inner Mbugu also had two lateral consonants; secondly, we will see that a number of the present-day lateral fricatives are relatively recent developments; and thirdly, that the Southern Cushitic languages are not the only source for the lateral fricatives. In Table 3.1,1 present various words containing the lateral fricative that have been rendered differently in earlier sources. The sources are: Archdeacon Farler (1885) (F85 in tables), Lieutenant Storch (1895) (S95), Mein­ hof (1906) (M06), Green (1958) (G58), Tucker and Bryan (1974) (T&B74) collected in 1959 and 1965, and my own data, labelled as (1994). Copland (1933-34) does not contain any relevant lexical items. Ehret (1980) is not deviant from my data. F85 kilte kulto Ita

S95 küsse kusso

06 kuhlo ta taxa

hulta vi-hilte mwata

muashla, muasa

mwasa mas ar­ èno su

matarinu tu senu eshla mgasho mushië sa usha shamiwa

esa sihe

G58 (le) kuhlo Ma hlaka muhla ihle kuthlea mwahla mahl ar­ eno Mu Menu ehla mgahlo mhlihe

T&B74 kuhlo hlá

mwáhla mahlarenu hlu 'éhla muhlíhe Ma

Mamilwa mbala m axaila kus a'amu

kisat

kisatu

kis a tu

kusamú, kuhla'amú kisatú, kihlatú

1994 kuhlé kuhló hlá Máka húhla 'ihlé kuhlea mwáhla

meaning long good plant give birth remember unripe, raw safari-ant fire

mahlarèno h;ú Menú éhla mgahló mhlihé hlá ahlú hlámuiwa mbahlu makahlu ku-sáme

clouds fall down tomorrow custom visitor moon open hair happy bushbuck dream foot

kihlatú

finger

Table 3.1: The development of the lateral fricative ł (hi) The first two sources are probably not reliable in their orthography: Farler presumably uses It for ł, but he has three items with t for ł. This may be a mistake, but it could also be that t developed into ł, see below. Storch has shl for ł but he also has ss, s and sh

54

CHAPTER 3

for words which now contain a lateral fricative. Given his German background, I take intervocalic double ss to indicate s as in kassa 'bad' (Farler has kusa and nowadays it is kusa), but see below. Probably, there was variation in pronounciation between s or sh and ł since Storch has muashla 'fire' versus muasa 'heat'. Meinhof was the first professional linguist to collect data on Ma'á and his data can be trusted to be phoneti­ cally accurate. His symbol for ł is s and his results correspond to ours with some in­ triguing exceptions. First, the s in visatu; second, his notation of / in mbala and maxala which correspond to present day í followed by u, for both of which í must be a later development, see below. Third, he noted two other sounds which do not occur in the present day language: z for β and t for tł The first one occurs only in the word izaka 'Köcher (quiver)' which is present-day iziaka in Normal Mbugu, from Pare iziaka, the 1Mb counterpart being isiye. Thus this once-noted occurrence of the voiced lateral is not in an Inner Mbugu word and we can draw no historical conclusions from it. The second one, t for tí, occurs in two words, 'to plant' and 'to give birth' which both contain ł nowadays and these two notations are historically important. Accepting the possibility of two distinct lateral sounds in 1Mb as it was spoken a century ago urges us to re-examine the sources before Meinhof in this light and this results in the following evidence for a former lateral affricate tł •

All authors before Tucker and Bryan have at least two different orthographic signs and these usages have some consistency: Farler's It = Storch ss = Mein­ hof 's t= Green's thl.4 • Farler, Meinhof and Green have a grapheme which includes t in combination with a lateral: It,t, thl. • The first seven items in Table 3.1 are those that probably once contained this lateral affricate. Three of these have a lateral affricate in their etymological source: kuhlé 'long' from Iraqw tleer (ku is added to adjectives), 'ihlé 'unripe, raw' from Iraqw ?eetl 'fresh', or from Yaaku 'ihle' 'pasture, grasses', and kuhlea 'safari-ant' from Sagala tlyeha 'white ants'; possibly the tly in Sagala was also an affricate.

The second group of words in Table 3.1 contain hl in all the reports on the language since Meinhof. The earliest sources were probably just confused. These lateral frica­ tives are surviving laterals and they can be traced by cognate words containing a lat­ eral fricative. Apart from the Southern Cushitic languages and Dahalo, the lateral fricative also occurs in the Taita Bantu languages Sagala and David, relatively close to the Mbugu area. In Davidaą, the voiced lateral fricative is a phonetic realisation of r in the Mbale dialect of Davida (Harris 1978, Philippson p.c.). In Sagala its occurrence is inferred from the notation tly by Wray (Ehret and Nurse 1981). I consider both Dahalo and Cushitic influence in Taita as representatives of Old Kenyan Cushitic and since we have argued above that an original Cushitic predecessor of Ma'á is best sought in Old Kenyan Cushitic too, the lateral fricative may have been inherited from the original Cushitic language.

LINGUISTIC HISTORY

Ma'á hlá hlohlóshi ihlarí hlá hlahá mahl aha igwahlú kikohlé óhlu mwáhla i'ohlú hieru (m)hlihé

55

meaning source ache Sagala: lya start Sagala: tlyoa cloud Sagala: tlyara stab, pierce Sagala: tlyai 'stab, sew' burn Dahalo: hlahh anger Dahalo: hlaka 'anger' cheek Dahalo: ngoohlo 'soft palate' instrument Iraqw: xoohla insult Iraqw: ? o o h l 'curse' Iraqw: ahla fire cheek Iraqw: ?unhla (West-Rift: ?untla) grow Iraqw: tleeruw 'become tall month Alagwa: hleehhee (West-Rift: hlehhen) Table 3.2: Surviving lateral fricatives

The sound hl is the most characteristic sound of Inner Mbugu. It is the sound that other people point out as an example of why 1Mb is so 'difficult'. Therefore it is not surprising that there are words for which 1Mb has developed this lateral sound from other alveolar consonants such as s, /, t, and c' from a variety of sources, even as a means to transform words from Normal to Inner Mbugu: the lateral fricative has ac­ quired the function of showing deviance. Interestingly, in a number of cases the de­ velopment of the lateral sound may have been stimulated by the existence of a similar root containing hl in Iraqw by way of convergence. Ma'á kihlatú makahlu ahlii mbahlu nhxwahlé mhla'aré hlí hláma hlanú mahloma hlámuiwa i'ahléta hlámu

meaning earlier attestation finger M06, S1895: kisatu dream M06: maxala his mother G58: ali'i bushbuck M06: mbala partridge runnel refuse leave behind pay fat, oil G58: hlamilwa joy room for goats sweet Table 3.3: Newly developed

source

Sagala: mbala NMb/Pare: nhkwaré NMb: mharé Iraqw: si' Maasai: a-lam Sagala: shana Oromo: c'ooma Fusion of Sh. tamilwa and Ir. hlaa' 'want'. Fusion of 1Mb ahla'ú and Ma. aleta Fusion of Sw. tamu and Dah. hlááhhame lateral fricatives

Two lexemes that developed a lateral fricative from s are now back to s: silu 'borrow'

56

CHAPTER 3

from Sagala sile 'debt' was hliu in Green (1958) and kusamé 'foot' was kuhla'amú in Tucker & Bryan (1974), but kusa'amu in Green (1958). There are some indications linking d and hi: Meinhof has Mara 'seize' next to dara; after him Green and all others have dara 'seize'; this is either a mistake by Meinhof or it is related to the equivalence of Southern Cushitic tl with Eastern Cushitic d, see Kießling & Mous (to appear). There is also variation between hl and r in éhla or èra 'any traditional cultural item'. Other words in Ma'á containing hl that were not discussed in this section are: hlásu 'ask'; kikahlé 'bangle, bracelet'; hlagha 'broad'; ihluhé 'cold (illness), catarrh, clan of Shambaa'; hlahla'á 'dance, play'; hle 'dawn'; bubuhlú 'gather'; hlané 'goat (male) '; hla'ári 'help'; buhle 'hide, skin (dried)'; kihló 'louse'; héhla 'meet'; ihlé 'name'; hid 'open'; hlutú 'pass on, let inherit'; kwahli 'path'; xehlú 'rat sp.'; lumuhlú 'rope'; hlúhlú 'slowly'; m-puhlé 'soil, sand'; hláhlámi 'sweep'; hlawi 'woo, ask hand in marriage'; hlagha 'win'; kohlijwa 'wet'. Concluding this section, we summarise that the lateral sound may have been in­ herited from a former Cushitic language, but, if so, it certainly became reinforced by the influence from Gorwaa/Iraqw and this latter influence is also the most probable force behind the second lateral (affricate) that must have existed in Ma'á in the recent past and that merged with the fricative. The lateral fricative was used as a sign of Ma'á identity and introduced into some lexical items. 3.4

Truncation rule The truncation process deletes the final syllable of a word. This rule has oper­ ated in the process of deriving Inner Mbugu forms from their Normal Mbugu counter­ parts: 1Mb té from NMb tenia 'cut', sú from sunga 'type of room', hú from hárúma 'shave, cut hair', 'ú from uvu 'force' and sako from s akame 'blood'. The operation of the rule is evidenced in various other ways: It is documented by comparison with the form of some lexical items in earlier sources, as for example in vasitá 'Shambaa' and kiMa'á 'Mbugu' which were vasitane and kimaanthi in Meinhof (1906), 'ó 'give' from ojo in Copland (1933/34), and hó'o 'vomit' from ho'omi in Green (1958). Sometimes there is present-day free variation between truncated and untruncated forms, as for example in kwahli or kwa 'path', M'ale or kú'a 'new'. There is dialect variation in the case of kechi 'your' in the Rangwi dialect (and in Green 1958), but ke in Magamba and Bumbuli. Full and truncated forms of the same root appear in derivation: u-ruká-o 'illness' derived from rú 'be ill' and mi-tém-o 'cuttings' (Meinhof 1906), derived from té 'cut', or derived from Normal Mbugu tema. This variation must have existed at earlier times since the documented history reveals cases of truncated forms that have not survived while their full untruncated bases did: mgirú 'man' was mgi in Storch (1895), minda 'house' was mi, mida, or minda, itorú 'spear' was ito or toro and i-hlari, ma-hlareno 'cloud' was also mahl in Meinhof (1906) (see section 3.7), while mkwàno 'skin (with hair attached)' was

LINGUISTIC HISTORY

57

mukwa, and ihóra 'hole in the ground" was iho in Green (1958), and still is ihó in the Bumbuli area. In addition there are Inner Mbugu forms that display truncation of the final syl­ lable from their original source, tú 'pound', úfu 'blow' from Iraqw(/Gorwaa) tu'-uut and 'ufaahh. In these last two cases the truncation rule makes consonant final verbs vowel final. The final syllable that is dropped from tu'uut happens to be a suffix in Iraqw/Gorwaa, and also in the case of sako from s akame what is dropped is identical in form to a non-productive suffix -mè in Inner Mbugu, see section 3.7. Thus in this case the truncation rule drops what can be recognised as a unit. In all other cases that we have presented sofar the truncated syllable did not coincide with a morphological unit. There is, however, a similar morphological effect to the truncation of the final syllable of words ending in éno, see section 3.7. All those Inner Mbugu forms that are based on the Normal Mbugu form by changing the final vowel into e and the tone pat­ tern to L(L-L)H can be seen as truncated words which once had an ending -éno, or other -éCV endings. The tone pattern L(L—L)H is a remnant of the tone pattern that these endings -éCV impose. For example the Normal Mbugu word for 'bee', ngila was ngiléno in Inner Mbugu in Meinhof's times; since those times truncation has applied yielding present-day ngilé. The same resultant process of changing the final vowel to é applied to Inner Mbugu forms from a variety of other source languages: mlamé 'lie' from Iraqw ¡ama, gimbé 'noise' from Gorwaa gimba, kikohlé 'instrument' from Iraqw/Gorwaa xoohla, and idié 'dog' from Mbugwe diyo or Maasai ol-dia. The process of changing the final vowel to é and imposing a L(L--L)H tone pat­ tern (or of adding éCV and subsequent truncation of the final syllable) affected some Inner Mbugu words in recent times. In the earliest literature on Mbugu, other final vowels appear where present-day Inner Mbugu has é: 1885/1903 miaga gomai magagala kusaamu ngila

Now mIage gomaé magalé kusamé ngilé

meaning woman cloths maize leg bee (but ngila in NMb)

Table 3.4: Development of final é This process of changing the final vowel to e after truncation and resulting in roots that are maximally disyllabic is not of the natural type of utterance final reduction, related to stress or tone. The rationale of the truncation rule is to deform. Truncation while changing the word final vowel to a particular value is not uncommon in lexical manipulation or in-group language (see, for example, Niceforo 1912:166). Ehret (1980:110) discusses this process as a regular historical change in pre-Ma'á. In his formulation the rule deletes all verb stem final and word final consonants and it is a

58

CHAPTER 3

relatively late rule, being number 45 in a total of 59 historical rules deriving Mbugu from Proto Southern Cushitic. Ehret suggests that the truncation rule is a device for rendering words vowel final, i.e. a mechanism for adjusting them to the Bantu frame and Thomason (1997:482) adheres to this view. I disagree with this explanation of the truncation rule for several reasons. One is that in many instances the input is already vowel-final and the rule truncates a complete syllable. A second is that the deletion of the final syllable does not lead to a form that is markedly more Bantu-like, but it does make the word different enough not to be like the Normal Mbugu equivalent. Thirdly, it also applies to Bantu words and in those cases there is no reason at all make the words more Bantu-like. In my view the truncation rule is a typical lexical manipula­ tion rule comparable to the truncation with final  in popular speech in Western Euro­ pean languages, e.g., demo. I thus do not consider the truncation rule as a suitable chronological landmark, contra Thomason (1997:482). 3.5

Remnants of non-Bantu grammar The elements of Cushitic grammar in Ma'á are crucial evidence for those who view Ma'a as the outcome of a gradual bantuisation process. Thomason is the most explicit defender of the gradual bantuisation thesis. It is therefore crucial to discuss this evidence in some detail and to investigate what historical conclusions can be drawn from it. The elements are all morphological in nature; Inner Mbugu syntax is identical to Normal Mbugu syntax. In such a discussion the following issues need to be addressed: • • • •

What What What What

counts as grammar? is the source for the deviant elements of grammar? counts as "Cushitic" grammar? is retention and what is innovation?

The first issue, what we should characterise as "grammar", has two sides. First of all, the distinction between grammar and lexicon is not immediately obvious. Take, for example, the closed sets of function words such as pronouns and demonstratives. Some people are inclined to subsume these under the heading grammar, while I con­ sider the forms of these function words as such to be lexical items whereas in my analysis agreement and (sub-)categorisation characteristics belong to grammar. For this reason, the closed sets of function words are discusssed in section 3.9. The second side of the label "grammar" is the issue of productivity. For each element discussed in this chapter the following factors of productivity are discussed: whether the element can be separated from the rest of the word, whether meaning can be attributed to it, whether it is parallel to a grammatical element in NMb, and its occurrence on roots of various sources. None of the non-Bantu elements are fully productive in the sense that their use can be predicted or that they are used with newly adopted words. In order to establish the source of a grammatical formative the search is for a language in which that formative is a productive affix and the occurrence of at least

LINGUISTIC HISTORY

59

one lexical item in 1Mb for which both the root and the formative come from that source language. Unfortunately, the candidates for non-Bantu grammar in Inner Mbugu are not straightforward and the establishment of the source becomes an issue. It should be obvious that only those structural elements that can be traced to a Cushitic language count as Cushitic; ''non-Bantu" has been assumed to equal "Cushitic" all too often. A second word of caution is required for the use of labels such as "Cushitic" and "Bantu"; clearly what is important for us here is not general Bantu characteristics but rather the characteristics of Bantu languages of the area and in particular those of Pare and Shambaa, and for "Cushitic" we should carefully distinguish between South­ ern and Eastern Cushitic. The crucial factor is whether the grammatical elements in question are reten­ tions or innovations. The gradual bantuisation hypothesis requires that the non-Bantu structural elements are from one and the same source and that this source is the origi­ nal (Cushitic) Ma'á language. Innovations of structural material, however, do occur both in the form of borrowing of grammatical elements as well as in the form of reanalysis of recurrent sequences as affixes in borrowed or original material. The latter is most easily recognisable as innovation. Any evidence for innovation or retention will be discussed for each of the candidate non-Bantu grammatical elements. When there is evidence for coexistence of Bantu and non-Bantu structures, this in itself is not necessarily proof of continuing Cushitic structure, as Thomason argues (1997:482). Instead, this coexistence could be the result of the use of non-Bantu features to con­ struct a deviant parallel lexicon, as might be the case with the nominal -éCV endings. There have been several propositions of Cushitic grammatical elements in Ma'á but some of these turn out to be: 1. Bantu, or 2. typological in nature and therefore not valid arguments, or 3. ghost grammar. Each of these will now be dealt with in turn. Falsely claimed Cushitic grammar that is in fact Bantu. The amplificative suffix -sha for nouns and verbs which according to Thomason (1983:214), following Ehret, is derived from Cushitic with no parallel in neighbouring Bantu, is in fact present in Shambaa (Roehl 1911:190-191) and Zigua, and developed from a common Bantu causative suffix -ish. The same development for the same suffix has also been ob­ served in Swahili (Ashton 1944:244-245). According to Thomason (1983: 215) Ma'á, uses the prepositions he and na, both from Southern Cushitic roots, after verbs of movement where Pare and Shambaa do not require insertion of a directional marker. The pro-clitic locative he- is not the Southern Cushitic *ha 'this' as in Iraqw hami 'now' (Ehret 1980:304) but the regular reflex of the Bantu locative class 16 in Pare and Inner and Normal Mbugu. Verbs of movement require a locative object: objects that are not locative can be made locative by the additicn of the class 16 prefix, (a normal Bantu procedure) or by the Bantu general prepostition na (see section 8.1). Ehret (1980:49) gives a supposedly Cushitic singulative -a as in támbala from tamba 'snail', but this is not a singular-plural pair and in fact nhtambala 'snail' (I+NMb) or also nhtamba (I+NMb) has Bantu sources such as Sagala kitambála

60

CHAPTER 3

'snail', Pare itamba 'snail', and Shambaa ntambala 'lentil sp.' The loss of/ intervocalically is a dialect rule in Shambaa and very common in East African Bantu in gen­ eral. According to Thomason (1983:210) the fact that possessives can be interpreted as suffixes is a specifically Southern Cushitic feature which is not paralleled in Bantu languages. However, possessives are only suffixes when attached to kin terms. The final vowel indicates 'my' vs 'his/her': 1Mb ii for third.person possessive is parallel and derived from NMb and Pare è for third person ((3.1)-(3.3)), just as -á indicates first person possessive in 1Mb, NMb and Pare ((3.4)). Thus, it is not a Cushitic rem­ nant but has a perfect parallel in Pare, and other Bantu languages. (3.1) ahlii 'his mother' [1Mb]. E.g., ahlii Charles 'mother of Charles'. Cf. Pare, Sagala mèe 'his mother' and compare also Pare and NMb i-mèe_ parallel to iahlii 'female domestic animal that has calved' [1Mb] (3.2) igii va-igii 'his brother/father' [1Mb]; [NMb: va-vée] (3.3) irjinhtii 'his/her sibling of opposite sex' [1Mb]; [NMb: rumbúryee] (3.4) irjinhtá 'my sibling of opposite sex' [1Mb]; [NMb: rambúrya] Thomason (1997) and (1983) reports that Ma'á cannot do without a copula in Noun - copula - Noun sentences where Pare and Shambaa can. This statement is not correct. The Pare nominal sentences that I elicited obligatorily contain a copula in such constructions. Note that Dahalo can do without the copula in Noun - Noun sen­ tences (Tosco 1991:89). Thus, there is no reason to assume that this property of Mbugu, both Inner and Normal Mbugu, is a remnant of Cushitic grammar. It has been claimed that the use of prepositions for locative phrases as opposed to the use of a locative suffix is a sign of Cushitic grammar. This is a misconception of Bantu grammar, since the use of locative suffixes to the exclusion of prepositions is not a common Bantu feature. Finally, the implosive pronunciation of voiced stops has been mentioned as a feature not attributable to Bantu languages, (Thomason 1983:203). Implosives as allophones of voiced stops are omni-present in the Bantu languages of the area: implosive pronounciation of voiced stops is the norm for coastal Swahili. Falsely claimed Cushitic grammar: typological and thus not valid. It has been claimed that Ma'á makes considerable use of adjectives and that this is a Cushitic rather than a Bantu feature. We show in chapter 7.1 that most Inner Mbugu (Ma'á) adjectives have Normal Mbugu parallels. Thus there is in fact no significant difference in the number of adjectives beteen the two languages. Furthermore, the general statement does not actually hold. The number of adjectives in Iraqw is not very much different from the number of adjectives in Bantu languages such as Chaga or Pare.

LINGUISTIC HISTORY

61

Another such typological claim is more cognitive in nature and in itself more promising for showing remnants of an earlier language situation. The claim is that Ma'á reflects the typical Southern Cushitic semantic pattern of a five-term colour sys­ tem, black-white-red-yellow-green, each expressed by a simple adjective, while Bantu languages have a three-term colour system, black-white-red. In my data in both Inner and Normal Mbugu there are two words for 'white', a verb and an adjective, an adjec­ tive for 'black' and an adjective for 'red'. The words that Ehret has for 'yellow' is -kenya 'young' and ma-kénya 'new grass' (Ehret 1980:251). In my data, the word -kénya 'fresh, unripe' is an adjective in the (shared) Normal Mbugu lexicon and thus there is no reason to look for Cushitic cognates. The Southern Cushitic correlate that Ehret proposes is the Iraqw adjective qany which is a cow-colour term and cow-colour terminology is not colour- but pattern-based and is much richer than a five-way divi­ sion. Ehref s word for 'green' is -háko related to ma-háko 'grass' — I only have the latter — which Ehret connects with Burunge hiqas 'to cook leafy greens' (Ehret 1980:306). This comparison cannot be maintained as in Burunge this verb is hiiqaas 'to cook meat (but probably also any other side-dish such as greens)' and derived from qaas 'to put', a semantic specialisation which has no connection with 'grass, greens' or 'green'. Falsely claimed Cushitic grammar: ghost grammar. Ehret sees a left-over of Cushitic gender marking in i'alú 'sheep' versus i'ale 'ram', but there is no other instance of sex differentiated this way and the Cushitic languages generally have two distinct roots for the sex difference in this animal. Instead, (V)alé is the truncated form of aléno, which has been reported too; both are based on the same source ('Other' Cushitic, cf. Oromo oolaa 'sheep'); once both forms existed side by side, after which a semantic differen­ tiation took place. For some of the other claims of Cushitic grammar in Ma'á I have no evidence at all in my data. Since no language material was given to substantiate these claims I cannot evaluate these differences in observation between me and Ehret (Thomason's source). It has been claimed that Ma'á can have a juxtaposition of possessor following the possessed, next to the option of a pronoun (see section 7.2).5 There is no sign of this in my data on Mbugu. In addition, Ehret reports on adjective forming suffixes -7 and -a of Southern Cushitic origin. I have no evidence for such suffixes (see section 3.9, Ehret 1980:Ch.5 67-69). Apart from the forms of the pronouns, possessives and demonstratives mentioned above and discussed in section 3.9, the claims for nonBantu grammatical elements are: 1. the non-agreement of some modifiers; 2 the insta­ bility of noun class prefixes; 3 a verb "to have"; 4. recurrent endings in the nouns and verbs. The non-agreement of some modifiers: Demonstratives show no agreement. Possessives show no agreement with the head noun when they immediately follow the noun but they do when they follow another modifier or when they are used predicatively. This agreement behaviour is deviant from the Bantu standard and also from

62

CHAPTER 3

Mbugu. It is not readily attributable to a Cushitic or Nilotic source, but is most likely a relic of code-switching patterns as we argue in some detail in section 3.10. The instability of certain noun class prefixes for certain nouns: This observation refers to the claim that the noun prefix of some words of non-Bantu origin in Inner Mbugu is optional. A possible explanation is that these words may have acquired their noun prefix at a later stage. Reference to the instability of the noun class system starts with Meinhof (1906:295), who remarks that he heard noun prefixes when eliciting but heard the same words from the same speakers without the noun prefix on other occa­ sions. Indeed, up to today, in the Bumbuli area where Meinhof did his research, the word for 'fire' is ahla, from Southern Cushitic ahla fire'. In the other areas, the word is mw-ahla with the noun prefix. Other words in Bumbuli in class 3 without the class prefix m(u)- are listed in (3.5). (3.5) hi'i'he (class 3) naró (class 3) haráza (class 3) (m-)lamé (class 3)

moon, month salt river lie

These nouns all show agreement in class 3 and have the noun class prefix mi- (class 4) in the plural. All other words in class 3 need a noun class prefix. A similar phenome­ non has been attested in a few words in class 14.1 lacking the noun class prefix vu-, but again showing full agreement with 14.1 (see (3.6)). (3.6) xaró (class 14.1) basà (class 14.1)

problems face

Meinhofs remarks refer to this phenomenon of prefixless nouns.6 This is still charac­ teristic of the Bumbuli area where Meinhof collected his material. Moreover the pre­ fixless words that occur in his data are the same exceptional lexical items that show this behaviour today, in other words, the "instability'' is stable It is important to note that the nouns in question show full agreement properties of the appropriate noun class. The explanation could be that the noun prefix is the last and the least salient feature in the parallel lexical addition. A verb "to have " in Inner Mbugu as equivalent to "be with" in Normal Mbugu: The Normal Mbugu (and general Bantu) construction to express possession is a sub­ ject prefix followed by the comitative or general puipose preposition na 'with', see section 8.1. The Inner Mbugu equivalent is a verb lo 'to have'. The origin of this verb is unclear. In this respect Inner Mbugu and Normal Mbugu are not parallel and the Inner Mbugu construction has been considered to be non-Bantu and Cushitic in nature. The two constructions are, however, not as different as they may seem at first sight. In Normal Mbugu the locative class 16 prefix he followed by the preposition na is used to express "there is" and Inner Mbugu does the same, using the class 16 he prefix with lo. In the "there is" construction lo is parallel to na. If we analyse na in Normal

LINGUISTIC HISTORY

63

Mbugu as an exceptional verb, parallelism between Normal Mbugu na and Inner Mbugu lo in both the possessive and the existential constructions is restored. For addi­ tional evidence that the speakers treat na as a verb, see section 8.1. It is true that Cushitic languages in general do not use a construction with a comitative preposition to express possession. However, I doubt whether there is a conceptual gap between Cushitic and Bantu languages in this respect because in a language such as Iraqw the verb "to have" is also used intransitively meaning "to be together".7 Recurrent endings in the nouns and verbs: The major candidates are the causa­ tive extensions and the detachable disyllabic noun endings containing a high toned e. These are now discussed in detail. 3.6

Ma 'a non-Bantu causatives The non-Bantu element in Ma'á that comes closest to productivity is the causa­ tive suffix -ti. This ending is clearly recognisable as a grammatical element for the following reasons: 1. the forms containing it are derived from a base that exists in the language8 2. most of these forms clearly have causative meaning9 3. the parallel forms in Mbugu contain the regular Bantu causative -ij,]0 which is also used for any newly formed causative in either of the languages (see section 5.6). In the following list of verbs with the causative extension -ti only 'agháti 'compare', gáléti 'bring', hokwáti 'be late', kweti 'quarrel' have no causative equivalent in Normal Mbugu (i.e. the equivalent is another lexical verb or is absent in my files) 4. there is evidence of this ending added to "foreign" material. causative 'agháti 'akúti waháti susughúti firíti gugulúti keti kiríti báléti héti dáráti tatí hláti laháti húti

meaning compare sprout water, let drink drive finish chase away lend return (tr) make move let arrive hold tied cheat shut lose (tr) fill

base 'aghá 'aku waha sosúghu

fi

gugula ké kiríi

bale hé dàra tà hlá Iaha hú

meaning compare shine, white drink go on finish run away put return (intr) move house arrive seize cheat open lose be full

source Pa: agh-an-a 'agree' IR: ?awaak 'white' IR: wah 'drink' IR: suruk 'move on'

OR: fit'a Dah: OR: gugad'd'a, also Bondei Dah: kaaj-; OR: kaa, keeta Dah: kiir- 'go and come back' 11 Dah: b'alak- 'move house' Dah: ha?'come' cf. Dah: d'ar- 'put together, collect' Dah: tabadi lie' cf. Dah: tl'aa?aj- 'let go' MA: a-laa 'be untied' MA: a-iput 'fill' with backforation on t as causative

64

causative 'óti 'alalúti 'umáti 'uti háláti gáléti hléti ku 'urúti kwetí hokwáti káti

CHAPTER 3

meaning take out wear out stop (tr) cure dry (tr) bring be late straighten quarrel be late kindle

base 'o 'álálu 'úma 'ú

hàla gale hlé ku 'úru kwé hókwa

meaning give sb. worn out stand get better dry leave dawn straight fight set(of sun)

source

Table 3.5:1Mb causatives in -ti and their sources The syllable preceding the ending -ti tends to have a high tone. Green (1958) has a variant with a prenasalised -ti in chi'inti as a causative of chi'i 'carry off, send'. The verbs that contain the suffix ti may have an alternative form with the pro­ ductive causative -ij, in which case there is often a difference in meaning: róti 'let wait' and róíja 'refuse somebody', both from the verb ró 'wait'; kúrúti wá 'let oxen cultivate' but kúrúja mhé 'let somebody cultivate' both from the verb kúru 'cultivate'. An additional causative extension -ij is reported with no additional meaning in 'o-ti-j-a 'take', hle-ti-ja 'be late'. A number of verbs with the causative extension -ti have ac­ quired a specialised meaning, such as hláti 'close' from hlá 'open', and hléti 'be late' from hlé 'get up'. The fact that the non-Bantu suffix -ti is added both to verb stems that have un­ dergone truncation and those that have not, is presented by Thomason as evidence that a non-Bantu grammar was available while speakers adopted the language to Bantu standards. The truncation rule is undoubtedly an important operation in Ma'á. However, as argued in 3.4, the truncation rule is not a sound change in the Neogrammarian sense, but rather a device to make the word Inner Mbugu. Thus, it may have been valid during the whole period of register building and therefore it can not be used as a chronological measurement. The second objection is that this ending ti is not simply a Cushitic causative suffix that survived, at least not in Ehret's reconstruction which Thomason follows. Ehret has argued that -ti comes from a Southern Cushitic suffix -t. This suffix expresses either a stative or a middle voice meaning and renders the verb intransitive. In Ehret's analysis this in itself is not a hindrance to propose it as a source for a causative suffix because he supposes a reanalysis after the truncation rule had op­ erated, that is, that a sequence of this middle voice suffix followed by a causative suffix , -t-iis, truncated to -ti, was reinterpreted as the new causative suffix, cf. Ehret (1980:63). Although this scenario cannot be ruled out, it is subject to doubt. The combination -t-iis does indeed exist in Southern Cushitic languages but it is very rare, while another com­ bination with a causative suffix, -m-iis is more common. Thus one would expect -mi

LINGUISTIC HISTORY

65

rather than -ti for the Ma'á causative if it came about along this path. Moreover, none of the 1Mb verbs in -ti has a Southern Cushitic cognate for which such a derivation -t-iis has been attested. If one accepts Ehret's analysis, as Thomason does, one cannot at the same time use it as evidence for persisting Cushitic grammar since such a reanalysis does not need any proficiency in a Cushitic grammar by the speakers who did the reanalysis. On the contrary such proficiency would more likely be a hindrance in using a de-transitivising formative for a causative. Thomason (1997) cites the example of sahe-ti from Ehret (1980) to show the ordering. The form sahe-ti 'to collect' is derived from sahe 'to meet' which according to Ehret is a reflex of Southern Cushitic t'ahhem. Thus the ti is added after truncation of the final consonant m (see Ehret 1980: 174, no. 9,10). Ehret's etymology cannot be maintained: Burunge cehheemis 'greet', or actu­ ally 'to pass on greetings' is the only cognate form to reconstruct Proto outhern Cushitic t'ahhem. The Burunge form, however, is derived from caahh 'get to know', which is the reflex of West-Rift Southern Cushitic tsaahh 'recognise, familiarise'. The semantics of 'to greet' in the derived Burunge verb are needed to connect it to 'to meet' in Ma'á; the form sahe in Ma'á requires us to assume that there was application of the truncation rule. Thus the causative -ti must have been a later addition, but not under the influence of still available Cushitic grammar as argued by Thomason, be­ cause a Cushitic suffix -ti as such never existed in Ehret's scenario; it is a Ma'á crea­ tion. If the etymologies are accepted, what the example shows is nothing but the fact that the ending ti with causative meaning at one point in history could be added to ex­ isting verbs; this in itself can be shown more convincingly by the addition of the suffix in agha-ti to a Normal Mbugu Bantu root (first example of Table 3.5). However, we could consider other sources for the element ti. Another conceiv­ able source of -ti is the Maasai causative prefix ati-. Semantically this is of course more satisfactory but it is placed at the other end of the verb. Moreover, it is limited to one class of verbs only and the one 1Mb verb with -ti for which a Maasai source can be proposed does not belong to this class. This proposal is not acceptable either. The best candidate source for the Inner Mbugu -ti is the Dahalo causative suffix -Vd which in conjugation is followed by a vowel, the most frequent vowel being i. It has a variant t. It also functions as verbalizer and is used in adaptation of loans. Ac­ cepting this source for the causative -ti implies a crucial role for Dahalo or rather its predecessor, that is, for a Cushitic language that has undergone the sound change of s to d/t for the causative. According to Tosco (1991:43) "[t]his stem extension corre­ sponds to the D reflex of the general EC [Eastern Cushitic] causative in {s}". The lan­ guages of the Iraqw group all have the s reflex of the Cushitic causative s. In Dahalo there is a homophonous suffix that functions as a frequentative. Among the verbs for which Dahalo evidence for an Old Kenyan Cushitic origin is proposed in the table above — and they are numerous — there is unfortunately not a single one for which an actual causative derived form in d is reported in Dahalo. This may, however, be due to the fact that the descriptions of Dahalo are not comprehensive and that its vocabu­ lary has not been fully investigated. Such derivations may very well exist since Tosco (1991:43) mentions the causatived as one of the few fully productive stem extensions.

66

CHAPTER 3

Apart from the causative function, an intensifying function can be observed in a number of Inner Mbugu verbs ending in -ti. Such a connection between Intensive and Causative also exists in the Normal Mbugu causative suffix -ish (see section 5.6). It would be interesting to compare the semantic conceptualisation in pairs of underived verbs and verbs in -ti with the equivalent verbs in the source languages to see whether this points to the cognitive pattern of a specific language. This is espe­ cially promising since some of the derivations that are semantically unexpected are completely parallel in both Inner and Normal Mbugu and are not found in Pare. Some of these less straightforward derivations in -ti have parallels in other — and different — languages, but more often than not these parallels are not very strong since the derivations are not causatives. For example: •

•

• • •

Both NMb and 1Mb derive "to close" from "to open" by a causative which is conceptually opaque. A relation between these two verbs is common in Bantu but with the derivation going in the other direction and the derivation being separative and not causative. In Inner and Normal Mbugu the verb "to be late" is derived by a causative from a verb "to dawn (of the sun)". In Bantu it is common for the verb "to be late" to be derived from a verb "to dawn (of the sun)", but with an applied + passive se­ ries of derivations rather than with a causative. In Inner and Normal Mbugu "to sprout" is derived with a causative from "to shine". In Maasai "to sprout" is derived from "to shine" but again with a differ­ ent derivation. In Inner and Normal Mbugu "to lend" is derived with a causative from "to put" and in Iraqw too, but again not by a causative. In Inner and Normal Mbugu "to take out" is derived from "to give" by a causa­ tive. I have not found a language in the area which parallels this (causative) derivation of "to take out" from "to give".

The conclusion must be that the causative is sometimes used to derive forms that can­ not easily be related to causative meaning and for which in other languages of the area other derivations are used. This suggests a wider range of functions for -ti and a regularisation of Normal Mbugu adjusting to Inner Mbugu. One might think that a multi­ tude of functions for -ti is precisely what Ehret's analysis predicts, given the fact that in his proposition -ti is a reanalysis of a truncated form of -ti plus another suffix. However, there are no other suffixes, apart from the causative, that are allowed to fol­ low ti according to the strict order restrictions of verbal derivational suffixes in South­ ern Cushitic. The influence of Inner Mbugu on the lexicalization pattern of Normal Mbugu could be interpreted as substrate influence after a shift to Pare/Normal Mbugu or as evidence for prolongued influence of the original Cushitic language in Thomason's scenario for gradual bantuisation. There are also other verbal endings in Ma'á of the form alveolar consonant + i (si, -ri, -li) which could be interpreted as verbal derivational suffixes, and as causa-

LINGUISTIC HISTORY

67

tives. A factor in the history of these "causative suffixes" may be that Bantu languages often have an unproductive "short" causative in i which invokes changes on the pre­ ceding consonant. The ending si is a clear causative in 'u'si 'cure' from 'w 'get better', gosi 'suckle' from gó 'suck' and not so clear in benesi 'die' possibly lexicalised from béni 'fail, miss' (cf. Oromo bada 'miss, fail'). It survives probably in the verb 'a'si 'ripen', and it is possibly also present in arési 'ask for', osi 'afraid', sisi 'laugh' (possibly derived from Pare and NMb seka 'laugh'), and kwasi 'overcome'. The suffix -si is a competi­ tor to -ti in the case of 'usi or 'uti (in itself additional evidence that the ending was causative in nature). Both forms, -si and -ti are equally close to the Dahalo causative -ti which contains a dental fricative. The ending ri is a clear causative in kwáhári 'make tired' from kwáha 'be tired' (cf. West-Rift Southern Cushitic kwahh-as 'rest' which contains a causative), dímíri 'drop' from dimi, dimu 'descend' (cf. Oromo dimmuu 'depth'). The causative kwahári is an optional variant of two other causatives from the same verb, kwáháti and kwáhája, without any discernible difference in meaning. The causative form of the verb fi 'finish' contains the causative element ri plus another causative element, either ti in firiti or ij in firííja, in both cases without making a difference in meaning. A fro­ zen form of this extension is possibly present in the following verbs but the meanings do not always force one to assume a causative interpretation: hiri 'hurt', kwári 'mix', shi'íri 'skin, strip bark of, hapári 'cover', cheri 'be early' and hla'ari 'help'. Only the equivalents of hiri, kwari, dimiri, and firiti are also causatives in Normal Mbugu. An ending li seems to be recognisable in fwáli 'scratch' (cf. u-fwá 'rash'), dodóli 'pick up' (from NMb and Shambaa dodola), and engéli 'get drunk' (from NMb and Pare engea). The causative character of this ending is not evident. It looks more like a suffix to render some Normal Mbugu verbs Inner Mbugu. There is evidence for a verbalising and loan verb adapting suffix i plus a penul­ timate high tone for which we have the following four examples: zeméri 'stay at the same height in walking' from zemera 'front', hlómi (v) 'anoint' from ma-hloma 'fat, oil', engéli 'get drunk' from NMb and Pare enge(l)a, dodóli 'pick up' derived from NMb (and Shambaa) dodola. This is possibly also how fwáli 'to scratch' derived from Oromo fala 'purge, clean' and gosi 'suckle' from Iraqw goso 'gap of incisor teeth' came into being. If the forms ufwá 'rash' and gó 'suck' developed through truncation, then si was possibly reinterpreted as a causative suffix through backformation. There is one indication for a suffix i which has the opposite function, i.e. nominalisation: ikúri (5/6) 'field' (IMb-archaic) is derived from the verb kuru, which has a parallel in Gweno. Nominalisation in -i is not uncommon for Bantu. Apart from the evidence for non-Bantu causative suffixes in Inner Mbugu con­ sisting of an alveolar consonant plus i, there are a number of other possible causative endings. Two Inner Mbugu verbs seem to be derived by a suffix -'a: gá'a 'kill, extin­ guish' from gå 'die' (cf. Iraqw/Gorwaa gwaa' 'die') and ka'a or xa'a 'put high, un­ cover' from ká ox xá 'get up, stand'. There are a number of verbs in 1Mb that end in a glottal stop followed by a copy of the preceding vowel: lá'a 'wait' (cf. Iraqw/Gorwaa

68

CHAPTER 3

lak-iit 'wait'), ti'i or chi'i 'carry off, send', pá'a 'cut, saw', xá'a or ha'a 'split, tear', hwá'a 'return', pu'ú 'shine', hó'o 'vomit, taste'. It is of course very speculative to as­ sume that these are all deponent verbs containing a suffix -'V, especially since some of them are intransitive. This formative is not productive in any respect. The vowel after the glottal stop may be necessessary for pronounceing the glottal stop. Note that Cop­ land (1933) has bu' 'go' for bu'u, ba' 'say' for ba'a and la 'wait' for la'a.12 Other simi­ lar possible causatives in Inner Mbugu are: hlúku 'drop' from hlú 'fall', bútu 'put into' from búu 'enter'. There are two more verbs which may contain such a causative ex­ tension: hlutú 'pass on, let inherit' (the Swahili equivalent kurithisha contains a causa­ tive), kuturu 'climb' (the Swahili equivalent kupandisha contains a causative). On the item, gá'a, Ehret (1980:263) remarks "Ma'á 'to kill' is an independent formation from that seen in the proto-Rift *gas-, by separate application of the causa­ tive to the root." In his reconstruction of Southern Cushitic verb extensions, Ehret gives Ma'á forms as reflexes of proto-Southern-Cushitic suffixes13 and in his analysis some have fused as a result of the deletion of final consonants in Ma'á (my truncation rule, see 3.4). Thus, in his analysis, the stative/ intransitive *-Vm has three reflexes: V, -m (in complexes of the form -mV), and -n (in complexes of the form -VntV), but the continuative *-Vt has -V as a reflex too, in addition to -t (in complexes of the form -tV and -VntV): and the causative *-Vs is represented by -V, and by -s (in complexes of the form -sV), and by -i (in complex -tí) (Ehret 1980:62-64). The consonant of any extension only survives if another extension once followed it and as a consequence Ehret assumes in a multitude of cases the presence at one time of a suffix that disap­ peared because of the truncation rule. The machinery that is thus built up results in a proliferation of non-Bantu grammatical suffixes in Ma'á, all made up to make the etymologies work. Given the fixed order of stative-continuative-causative in Southern Cushitic, the 1Mb -tV derives from *-VtVs in this analysis. Ehret derives 1Mb -si from the sequence causative-consequentive. However, none of the present-day Southern Cushitic languages shows evidence of allowing such a sequence. 3.7

Non-Bantu nominal endings Inner Mbugu contains some nouns that end in éCo or éCa, -éno, -éto, -éta, -éna, -era, -.14 None of these function as nominal suffixes in the surrounding or other Bantu languages. These formatives can be recognised as units for various reasons. One indication is the variation within 1Mb, especially present in the Bumbuli area: itabéno or itambi 'branch', kishambéno or kishambi 'sheep's tail', bughéno or bughu 'fear', ixeréno or ikeri 'earring', ihlaréno or ihlari 'cloud', nihéna or nihi 'meat', mkadéma or ikádo 'ten'. Another indication is the use of some of these suffixes in derivations which turn a NMb noun into an 1Mb form: ilambéno from ilámba 'banana-bark', ikindéno from ikinda 'banana-tree', mseséno from msese 'pullet', ijindéra from ijindo 'root'. The suffix -éno also appears in three words in Inner Mbugu for which no form without the ending has been recorded: kidighéno 'entry', mduduxéno 'calabash for oil', ximeno 'bird*.

69

LINGUISTIC HISTORY

In none of the cases can any meaning be attached to these formatives. I cannot confirm the collective meaning that was reported by Ehret, as quoted in Thomason (1983: 209), but this may be due to recent loss since Meinhof (1906) has balaxe 'bite' next to balaxeno 'two bites at the same time'. The only sign of productivity that some of these formatives, -eno, -eto, -era and -ere, show is that they can be added to imported (Bantu) nouns (see (3.7)) as is the case with" some of the recent banana terminology from Shambaa. For these words, -eno serves to make an Inner Mbugu parallel to the Normal Mbugu lexeme, see sec­ tion 2.9. These endings are also found with a number of words in Inner Mbugu that are not of Bantu origin. (3.7) Endings on 1Mb words ilambeno ikindéno mseséno ueto (M) mu-Meto (M) nhkaleto (M; ijindéra msésere

from a Bantu source banana-bark' from banana-tree' from from pullet' fig tree' from soot' from crab' from root' from from pulet'

Sh: kilamba Pa: nhkinda Pa: msésé Sw: mkuyu Sh: mídale Sh: nhkala Bantu Pa: msésé

The origin of these endings is difficult to ascertain. Formally, they look like Cushitic plural suffixes since plural suffixes of the shape -eeCV are common in Cushitic. And indeed the non-Bantu sources in (3.8) are mostly Cushitic, from both Cushitic sources. However, I have not been able to find a Cushitic source language with such a nominal plural ending.15 Below we suggest that Maasai is an other option for the source of -éta. (3.8) Endings on lu-'iréno m'aléta ilameto nihéna mkadéma aféta i'ahléta buxeta

1Mb words from non-Bantu origin 'sleep' cf. Oromo irriwa 'girl' cf. Oromo intala cf. Kikuyu mwiretu 'broom' cf. Oromo lammat'ee 'kind of stick' 'meat' cf. Oromo nafa 'body' 'ten' cf. Oromo knd'-aani 'door' from Iraqw afee 'mouth (pl)' 'kitchen' from Iraqw mw-ahla 'fire' and Maasai ol-aleta 'places where small animals are kept' 'rainy from Iraqw buhare (Meinhof 1906; contemporary season' Ma'á has bure)

A parallel case of borrowing a presumably Cushitic plural suffix of a similar shape can be found in Ilwana, a Bantu language with considerable Cushitic influence. Ilwana has borrowed a suffix -ena to indicate plurality of nouns in the class pair 9/10 where plu­ rality is not overtly marked, (Nurse 1994, 2000). The dominant Cushitic factor in II-

70

CHAPTER 3

wana is Orma (Eastern Cushitic), but there is no plural suffix -eena in present-day Orma (Stroomer 1987). Nurse (2000) suggests that the origin of -eena could be Dahalo. The suffix -era also occurs in Ilwana as a nominal suffix of unknown but pre­ sumably Cushitic origin (see Nurse 2000:124-132). The origin of these endings could lie in the reinterpretation of lexical material. A case in point is the word Vahléta for 'kitchen, part of the house where smallstock is kept' which is a fusion of Iraqw ahla 'fire' and Maasai ol-ale, pi: il-aleta 'demarcated area, calf pen' (Mol 1996). It could have been the case that Maasai plurals in -o and -a on words ending in et (for example those that contain the instrumental nominalising suffix -et) underwent reinterpretation as containing an ending eto or eta. The numerous nouns ending in é in Inner Mbugu probably go back to truncated endings in éCo or éCa. This loss occurred partly during the recorded history of 1Mb. The most frequent ending is éno. This observation and the fact that -éno has been added to very new concepts such as banana terminology show that this ending was once productive as a Ma'á-isation device. A similar case can be made for -éto since it is found on a recent borrowing from Swahili, e.g. mkuy-eto 'fig tree' in Meinhof s days (mkuyu in present-day Inner and Normal Mbugu). A number of Inner Mbugu nouns are derived from 1Mb verbs by a suffix mé and placement into a noun class: ki-kuru-mé 'agriculture' from kuru 'cultivate', ki-o-mé 'language' from yo 'speak', i-hari-mé 'whetstone, duster' from haré 'sharpen, dust' and u-té-mé 'wilderness' possibly from te 'cut' (but cf. NMb terna 'to cut'). There are three nouns containing mé that cannot be shown to be derived but the historical sources for these words are verbs to which this mé is added: Iw-ata-mé 'light' possibly from Iraqw gwa?aat 'be light', i-riti-mé 'steppingstone' possibly from Dahalo rat 'walk', and lu'ara-mé 'broad' from Maasai a-lala 'be wide'.16 The suffix is not pro­ ductive. Structurally this is how nominalisations in Bantu take place but the suffix -mé is not known from Bantu languages in the area. There exist infinitival suffixes me in Dahalo (Tosco 1991:59) and imee in Alagwa and Burunge but not in Iraqw. There is one instance that suggests a nominalising suffix -mu, i.e. ma-kwaha-mú 'tiredness' (1Mb) from kwaha 'be tired'. A possible source for this suffix is the Dahalo infinitival suffix -amu (Mous 1993: 79) although the tone would be deviant. Other 1Mb nouns ending in mú are luximú '(finger-) nail', (lu)-harimú 'horn', i'asemú 'breast' from Iraqw iseemo and the adjective kumumú 'short, narrow'. For these words there is, however, no evidence that mú is not part of the root. Ilwana has a nominal suffix -imu of unknown but presumably Cushitic origin (see Nurse 2000:124-132). The variation between nhtunguri and nhtungo 'small medical gourd' (1Mb) could either be taken as evidence for the truncation rule or for a non-productive suffix -uri for which there is no other evidence. Several Inner Mbugu words end in áo. Interestingly some of these appear as ei­ ther ending in àà or áo in the speech of one of Tucker's informants (dialect uncertain) (Tucker & Bryan, 1974). These words are mruxáa, mruxáo 'ill person' and ixutáa, ixutáo 'hoe'. 17 The endings of these words in their sources, predominantly OromoDahalo, are either aa or ooluu and for these same words Ilwana and Sagala show evi-

LINGUISTIC HISTORY

71

dence of the same ending ao or au, notably rao and itiráo. The phenomenon seems to be a relic of the Old Kenyan Cushitic connection, possibly related to Oromo gender marking: aa for masculine and oo for feminine and plural.18 Ma'á urukào ihutáo itiráo raó laxáo diláo iharáo anhkáoanhkáo ba'áo

meaning illness hoe liver arrow for bleed­ ing cattle bile king calabash-half carefully bangle (arm or leg. female)

Old Kenyan Cushitic source cf. Dahalo: ruk'o 'illness'; cf. ru 'get sick' cf. Oromo: hordaa 'digging stick' cf. Oromo: tiruu 'liver', Ilwana tir au 'liver' cf. Sagala: kirao 'barb of arrow'; Oromo: la'aa 'arrow' Oromo: rakoo [raxoo] 'an animal that is slaughtered at a marriage ceremony' (link: bile is used in rituals) Oromo: d'iira 'man, husband, male'

Table 3.6: Ma a words ending in ao As for the nominal endings I should like to conclude that where Thomason sees com­ petition between old and innovative forms as evidence for Cushitic structures being replaced by Bantu structures, I claim that these competing forms are in fact part of the double lexicon structure using endings at least some of which arose from a reanalysis of "borrowed" lexical material. 3.8

Locatives and other non-Bantu frozen grammatical elements The Inner Mbugu locative clitic -za for 'inside' is parallel to the NMb and Bantu locative suffix -eni as in (3.9) and (3.10). Example (3.12) gives two additional 1Mb words that happen to end in za but for which there is no evidence for a locative suffix. Example (3.11) shows variation of the same lexical unit with and without the clitic -za. (3.9) ahlazá(\6) 'kitchen' [1Mb] Derived from ahla and the locative clitic za whose NMb parallel -eni is present in the parallel NMb item ndikw-eni. (3.10) (m)haráza (3/4) 'river area' [1Mb] The NMb equivalent mwet-eni 'river area' (NMb) contains the locative suffix -eni to which the 1Mb -za is parallel. (3.1 \)ngwéza (9/10) 'bushes, wilderness'. Also ngwe.

72

CHAPTER 3

(3.12) itúngúza (5/6) 'honey-comb' [1Mb] mgerezá (3/4) 'spinach sp.' [1Mb] The origin of the suffix is the Dahalo locative suffix da (Tosco 1991:75-77), which is [ða] intervocalically. The voiced dental fricative (also present in Arabic loans in Swa­ hili) is generally rendered by z in Tanzania, but in Farler's days the original pronuncia­ tion was still recorded, maratha [1Mb] 'river' (Farler 1885). There are relics of another locative suffix in Inner Mbugu, -'ú, namely in the word ahla'ú (9/10) 'room for goats' (1Mb) which is derived from áhla 'fire'. The par­ allel NMb word nynmb-eni contains a locative suffix -eni.19 Many of the Inner Mbugu adjectives begin in ku which probably goes back to the Cushitic gender marker for masculine (and neuter/plural) nouns. The various forms for the adjective 'new', ku'ale, ti'ale, ku'a, show variation between ku and ti, which is the Cushitic feminine marker. Iraqw adjectives have a preceding element ku for mas­ culine nouns when the adjective is used predicatively.20 Ehret (1980), however, gives these adjectives without the initial ku. The source of the ku is most likely the Cushitic masculine/neuter agreement marker ku fossilised onto the adjective. Within Bantu grammar words starting with ku- could taken to be in noun class 15 which typically contains words such as kumoso 'left' and kuúmé 'right'. 3.9

The origin of f unction words All the closed sets of function words: possessives, demonstratives, personal pronouns, and question words have an Inner Mbugu form next to a Normal Mbugu form. The categorisation is in all these cases identical to Pare or other Bantu lan­ guages. The forms, however, are from other languages. Their origin is difficult to dis­ cern due to adaptation, truncation and paradigmatic changes but most indications point towards Cushitic, and in particular towards the Old Kenyan Cushitic source. The Inner Mbugu demonstratives are monosyllabic, invariable and with high tone. There is some dialect variation. The demonstratives in Bumbuli end in 'ú. It is possible that these longer forms are original and that the demonstratives in the other areas underwent truncation (see section 3.4). Given the homogeneity of the ending 'ú in Bumbuli, it is also possible that this ending 'ú is the remnant of a grammatical marker. Demonstratives without this ending occur in Bumbuli too. My data are insuf­ ficient to determine a grammatical function of the presence or absence of this ending 'ú. See section 3.8 for a locative ending 'u. IMb-Bumbuli yá'ú ka'ú hú 'ú

1Mb yà ká hú

meaning this that that (far)

Table 3.7: Inner Mbugu demonstratives The fact that the demonstratives distinguish three degrees in distance is a Bantu phe-

73

LINGUISTIC HISTORY

nomenon. Iraqw (Southern Cushitic in general) and Maasai have a four-way distinc­ tion while Oromo and Dahalo have a two way distinction. In Bantu and in Cushitic the demonstratives follow the head noun. In Bantu languages the demonstratives are inde­ pendent words, as in Inner and Normal Mbugu. In Iraqw and Dahalo demonstratives can be suffixed. Inner Mbugu is deviant from all neighbouring languages in the fact that demonstratives show no agreement with the head noun. The noun class systems of Cushitic and Bantu are completely different and a complete breakdown in agreement is not uncommon in situations where languages meet that have different noun class systems, even in situations where such differences are minor as in the case of Lingala (Motingea Mangulu 1996). The forms of the 'demonstratives are clearly not Bantu or Maasai; in Table 3.8 and Table 3.9 non-Bantu parallels are given. The initial vowel of the Dahalo forms in Table 3.9, which are gender sensitive, is added when the demon­ stratives are used independently in Dahalo; ku is the agreement marker for masculine and neuter/plural nouns and t for feminine nouns; the final a and  indicate near and far deixis, respectively. This  for far deixis in Dahalo could be related to the Inner Mbugu form for far deixis. The Inner Mbugu form for near deixis ends in a like the one in Dahalo, and the second demonstrative in Inner Mbugu resembles the third one in Iraqw or the masculine first one in Oromo, again assuming truncation. But, in fact, the source language(s) for the Inner Mbugu demonstrative remain(s) a puzzle.

Maasai Iraqw

DEM1

DEM2

DEM3

DEM4

ele -í

elde -sí(ng)

il -qá'

lido -dá'

Table 3.8: Maasai and Iraqw demonstratives

Dahalo Oromo

DEMI

DEM2

('u) kw a, ('i) ta kana, tana'

('ú)ku, Ci) tu suni

Tabte 3.9: Dahalo and Oromo demonstratives Table 3.10 compares the Inner Mbugu personal pronouns with those of the relevant Cushitic languages. For the third person Inner Mbugu uses the distal demonstrative. Though included in the table, the Inner Mbugu third person should not be compared with the other third person pronouns. The table shows that the Inner Mbugu pronouns do "look Cushitic" and in fact the closest resemblance is with the Dahalo forms, be it that the second person plural seems to derive from the column for the bound pronouns (and not the independent pronouns) and that we have to assume a paradigmatic level­ ling in Inner Mbugu adjusting the final vowel of the second person to that of the first person. Inner Mbugu follows the grammatical system of Normal Mbugu in that it does not distinguish sex in the pronouns. All Cushitic languages distinguish female and male in at least one person.

74 meaning l SG 2SG.M/F 3SG.M/F 1 PL 2PL.M/F 3.PL

CHAPTER 3

1Mb ání åri (h) ú

n(í)ne kúné (h)ú

Iraqw 'aníng kúung / kíing 'inós 'atèn kuungá' 'ino'ín

Dahalo 'ányi 'ááta 'údn/'ídi nyányi 'atta 'iimmámn

Dahalo reduced 'an 'u/'i +G 21 nyan

Dahalo--bound 'i

ku 1 ki du I di ni

kunná/ kinná inya

Oromo ('a)naa sii 'isa, 'ishii nuu 'isani 'isaani

Table 3.10:1Mb personal pronouns and their Cushitic equivalents Table "3.11 presents the possessives in the relevant Cushitic languages in their inde­ pendent form. It should be kept in mind that the possessives in Inner Mbugu are inde­ pendent words and show agreement only in certain syntactic environments. If we as­ sume the Bumbuli first person possessive singular form to be original within Ma'á (1Mb)22, then the generalisation holds that all possessives begin in k. This in itself points to a Cushitic origin since the independent possessives in Cushitic languages begin with a gender marker and the most general one for masculine and neuter/plural begins in k. This is most regular in Iraqw and not valid at all in Dahalo. The Inner Mbugu possessives may look more Cushitic than anything else in form, but one Cushitic source is not easily singled out; Iraqw (or Gorwaa) comes closest. The Maasai pronouns are very different: lai lSG, lino 2SG, lenye 3SG, laŋ lPL, linyi 2PL, lenye 3PL. Inner Mbugu follows Normal Mbugu in not distinguishing between male and female possessors. 1Mb lSG . ghó 2SG ké 3SG kú ' lPL kánu 2PL kághu 3PL kini

IMb-Bumbuli ' kéchi = IMb = IMb = IMb = IMb

Iraqw kwe'ée' kók kós korén kohúng

Dahalo

Oromo

'ittsi

kiy(y)a ku'u/ki'i (kan)kee su'u/si'i isaa/isii -nyi'i keenna - 'keesani -'inyi'i isaani or worraa

'ín Table 3.11:1Mb possessives and their Cushitic equivalents

3.10 Shift or gradual bantuisation. Thomason (1997) proposes a long process of gradual bantuisation prior to a more recent lexical manipulation.23 Previously I argued against this (Mous 1996). I am now convinced that prior to parallel lexicon building a long process of gradual ban­ tuisation in a situation of language maintenance under bilingualism is most likely to have taken place as it can explain some of the agreement pohenomena in Ma'á as I show below. However, I also argue here that eventually this process of bantuisation must have led to a shift to Normal Mbugu and this shift opened the way for extension of the parallel lexicon through lexical manipulation. In Thomason and Kaufman's (1988) framework on historical change in a social context, mixed languages and in

LINGUISTIC HISTORY

75

particular Ma'á play a crucial role. The authors distinguish language maintenance ver­ sus language shift. Within language maintenance they propose a borrowing scale with five different stages. A sixth stage shows a break in transmission, which is posited for mixed languages. Within language shift Thomason and Kaufman distinguish normal shift and abrupt creolisation. For them Ma'á serves as an example of grammatical bor­ rowing. Let me examine this claim for Ma'á: According to Thomason and Kaufman, inflectional and derivational morphology, conceptual subdivision of pronominals and demonstratives, as well as syntax would all be borrowed from Pare. This would be such an exceptional and spectacular case of borrowing that the underlying prerequi­ sites need to be firmly proven. These are that the people once spoke another (pre­ sumably Cushitic) language and that they did not shift to a Bantu language. The first part of this prerequisite is likely on several grounds (see section 2.8). The second ele­ ment of the prerequisite, that they did not shift, is the issue at stake here. Thomason's view entails that as well as having a perfect command of this Bantu language, they kept their original Cushitic language which was gradually influenced by the Bantu language. If one accepts this scenario, it is important to keep in mind that this gradual grammatical borrowing would have occurred in a situation of prolonged bilingualism of two languages that are genetically and typologically far apart, whereby the people felt a strong pressure not to give up their original language. The evidence that Thomason (1997) puts forward for incremental bantuisation consists of two types: 1. there are non-Bantu grammatical features which are fossilised forms of the earlier stage of the Cushitic language, 2. chronologically ordered changes show that Bantu structural interference appeared at different times. The second argu­ ment does not hold in my view because the sound change that serves as evidence for this chronology is not the "regular Bantu-induced sound change that opened syllables by deleting syllable-final consonants" (Thomason 1997:482). In my analysis, this rule has a different form, a different motivation and is of a different nature. The truncation rule deletes the final syllable and not only the final consonant, see 3.4. As a conse­ quence, the motivation of the rule is not to open up syllables. It is true that Mbugu phonotactics requires open final syllables but most languages in the area are not mark­ edly different in that respect. Indeed the vast majority of Iraqw nouns end in a vowel too. The standard procedure in Bantu languages for opening up syllables is by addition of a vowel and not by deletion of the final syllable. The rule can therefore not be char­ acterised as "Bantu-induced". The truncation rule is not a regular sound change oper­ ating on Cushitic elements, rather it is a modification belonging to the process of par­ allel lexicon building, operating on a limited number of Bantu and non-Bantu lexical items, see 3.4. It is a truncation rule of the type we have seen in several manifestations of conscious lexical manipulation (Mous 2001a), applying to a number of lexical items; it is not an exceptionless sound change in the Neogrammarian sense. Some In­ ner Mbugu verbs will be truncated, others will not. The fact that endings such as -eto can be added to Bantu stems, as in 1Mb mkuyéto 'fig tree' (Meinhof 1906) from Swa­ hili mkuyu does not show that Cushitic morphology is active on recent Swahili loans;

76

CHAPTER 3

rather it shows that these endings are semantically empty and serve only to render words Inner Mbugu. A scenario of gradual bantuisation involves the borrowing of a noun class sys­ tem. To the best of my knowledge, the systematic addition of noun-class prefixes is something that has not been attested for any Bantu language in a borrowing context (and there are many), at least not the addition of CV- shape prefixes that we are deal­ ing with in the case of Ma'á. The strategy that Bantu languages use when borrowing nouns is that borrowed nouns are assigned to the noun class that has zero as an allomorph, often class 9 or 5, and in some languages to class 5 if the class 5 prefix con­ sists of a vowel (not CV-) i- only; another observed pattern is to reinterpret the first syllable as a prefix and assign the foreign element to the according class, for example Arabic kitáb becomes ki-tabu 'book', plural vi-tabu in Swahili. Let us imagine what it would take to borrow a Bantu noun class system. The present Ma'á noun class system is a typical Bantu noun class system. A noun class system has the following components: a noun prefix, a plural form in a specific class with its own noun prefix, lexical or non-semantic distribution over about sixteen classes, obligatory subject agreement and optional object agreement on the sentential level, i.e. on the verb, agreement within the noun phrase, the associative or genitive for example (as in (3.17), (3.18)). There are however some deviations in the Ma'á case: Demonstratives do not show agreement (compare (3.13) and (3.14)), and posses­ sives show only sentential agreement but no agreement within the noun phrase (com­ pare (3.13) and (3.16)). The possessive and demonstrative roots in Ma'á have forms that are similar to those in Cushitic languages. (3.13)ki-sima yá ni ki-kánu 7-well this is 7-our 'This well is ours.'

(Ma'á)

(3.14)i-síma ili ni 5-well 5:deml is 'This well is ours.'

(Mbugu)

l-étu 5-our

(3.15)ki-gi yá ni chá-shaba (Ma'á) 7:of-copper 7-thing this is 'This thing is of copper.' 3.16) nyamálo ghó 9.work my 'my work'

(Ma'á)

ku-úme k-ángu 15-right 15-my 'my right (arm)'

(3.17)mbúji ya-mgheni yáafwá (Mbugu) 9.goat 9:of-guest 9:FPST:die 'The guest's goat has died.'

(Mbugu)

LINGUISTIC HISTORY

77

(3.18) te-hé-lo kigí ki-háli chá u-murú-kúru (Ma'á) NEG-16-have thing 7-other 7:of 2SG-can-cultivate 'There is nothing else that you happen to cultivate?* In the subsequent discussion we take both Oromo and Dahalo into account as frames of reference for Old Kenyan Cushitic to which I assume the original language of the Mbugu belonged. Dahalo fits better historically but has undergone radical changes in its history. I concentrate on a scenario for the borrowing of a noun class system. The Ma'á noun class system has been claimed to show additional discrepancies and this again would be evidence for a recent or incomplete acquisition of the noun class system. The features that have been claimed are: (1) the noun class prefixes can be dropped, see for example Greenberg (1999:629); (2) a sizeable number of Inner Mbugu nouns lack the expected noun-class prefix (Thomason 1997:474). These two features, however, are not valid. The first feature, that noun prefixes can be dropped across the board is a hoax going back to Meinhof (1906), the details of which were discussed in 3.5. The second feature is based on a false impression. Thomason (1997:474) gives a number of 134 out of 373 nouns without a prefix, based on a count of Ehret (1980). These are alarming figures, because they suggest a high portion of nouns that do not have a noun class prefix (and possibly are not parallel to a Normal Mbugu noun), or else a high portion of nouns in class 9/10 where zero is an allomorph of the noun prefix and a tilted partition within this class favouring Inner Mbugu class 9/10 nouns with zero-allomorphs. My data show less dramatic figures: a low number of nouns that are not parallel in noun class, a relatively low to normal proportion of class 9/10 nouns and, within the class 9/10 nouns, a confirmation of a higher propor­ tion of prefixless nouns in 1Mb when compared to NMb. On a total of 884 parallel Inner and Normal Mbugu nouns, I have 128 that have no prefix. Twenty-five of these belong to class la which regularly has no prefix in either language (and in Bantu in general). The remaining 103 prefixless nouns belong to class 9 (and/or 10). Instead of 36% prefixless nouns, I have 11% prefixless nouns. Such a percentage is normal for a Bantu language; for example, Thilo Schadeberg (p.c.) informs me that Makhuwa (Tanzania/Mozambique) has 20%. The discrepancies between Thomason's and my calculation are due to several factors: Ehret does not mark morpheme boundaries, so possibly not all prefixes were recognised as such, nor does he write the voiceless nasal in class 9, thus rendering these words prefixless. Another factor is the fact that Ehret (1980) gives only those words that he can relate to a Cushitic source and there is in­ deed an effect that more class 9/10 words have zero as an allomorph of the noun pre­ fix, that is, 1Mb class 9/10 nouns are more often prefixless than NMb nouns. The important observation here is that there is indeed a considerable number (32) of 1Mb nouns in 9/10 that are prefixless while their NMb counterparts have a na­ sal prefix, e.g., 1Mb róro (9) = NMb nyika 'low plains', while the reverse does not occur (0 cases). This shows that the addition of noun prefixes does not always happen in class 9/10 in 1Mb. This, however, is not surprising given the fact that zero is an al­ lomorph of the class 9/10 prefix and that the non-zero allomorph of the class 9/10 pre-

78

CHAPTER 3

fix fuses with the root initial consonants and is therefore difficult to recognise as a pre­ fix for the language user. The 1Mb noun is in the same agreement class as the parallel NMb noun, the appropriate noun prefix is added to it. The allomorph of the added noun prefix depends on the stem and need not be identical to the allomorph of that noun prefix in the NMb parallel; thus, class 9/10 nouns in 1Mb choose the zero allo­ morph of the noun prefix if their initial consonant is a fricative. In fact, there are no 1Mb 9/10 nouns with a nasal added to a non-Bantu source, with the possible exception of nuŋa (9/10) 'nose' (if taken from from Iraqw dunga' or West-Rift 'urung- (but see section 2.8). Those that begin in a nasal already do so in the source, e.g., minda (9/10) (1Mb) 'house' cognate with Oromo mana or mina, and there is a small number of class 9/10 nouns in Ma'á that have a zero allomorph in class 9/10 where the regular allomorphy requires a nasal. Normal Mbugu also has a number of irregular, lexically con­ ditioned, zero allomorphs in class 9/10 (see section 6.2). Let us hypothesize how the borrowing of a noun class system could take place. The Cushitic nominal morphology and NP syntax have some similarities with Bantu both having Head-Modifier order — but many more differences. In Cushitic, nouns have covert gender with only three exponents, feminine/masculine/plural; the Bantu equivalent of gender is the noun class system, which is overt, has a large number of exponents and is likewise not semantically motivated. The structure of agreement in possessives and demonstratives is quite similar in Cushitic and Bantu and in this re­ spect there is no immediate explanation for the deviant agreement pattern in Ma'á. Possessives and demonstratives show agreement with the head noun both in Dahalo and Oromo ((3.19), (3.20)) both attributively and predicatively, as they do in Bantu. In Dahalo the demonstrative and possessive pronouns used predicatively are formed by a vowel plus the respective suffix; in Boraana, both attributively and predicatively, the same independent words are used. The genitive construction in both Dahalo and Oromo is formed by suffixation of an non-agreeing genitive marker to the modifier ((3.21), (3.22)). Thus the genitive construction in Cushitic is considerably different from that in Bantu, and the Ma'á genitive or associative ((3.17), (3.18)) is identical to the Mbugu (Bantu) associative. (3.19) nat'á-su'á-to (Dahalo, Tosco 1991:39) woman-his.F-that.F 'that woman of his.' (cf. nat'o 'women') (3.20) harka kana: obboi-eettii tiya (Boraana. Stroomer 1995:64,69) hand this.M; sister-F my.F 'this hand"; 'my sister" (cf. obboJaa 'brothers') (3.21 ) nat'a mfaüumé-éni-to (Dahalo, Tosco 1991:81) woman king-GEN-that.F the king's wife*

LINGUISTIC HISTORY

79

(3.22)mina ciita-a (Boraana, Stroomer 1995: 97) house chief-GEN 'the chief's house' Cushitic languages tend to have several number (singulative and plural) suffixes, none of which is still used as such in Ma'á. We can assume that they gradually got eroded due to phonological reduction in word final position and to a loss of functionality given the setting of language loss (despite intentions to keep the language). An addi­ tional factor in the loss of functionality is that in Boraana (and other Eastern Cushitic languages) plural forms are not obligatory when the plural semantics is obvious from the context. The remnants of some non-Bantu number suffixes survive in stem varia­ tions in Ma'á with no meaning difference, e.g., kishambéno ~ kishambi 'sheep's tail', bughéno ~ bughu 'fear', nihéna~ nihí 'meat'. The agreement system must have developed before the actual noun class pre­ fixes were added. Possibly the few nouns that lack a CV prefix are a testimony of this order of developments. The noun class agreement in Bantu is both within the noun phrase and outside the noun phrase, at sentence level. In a development of agreement it is most likely that it starts where the connection is tightest, i.e. in the noun phrase, as is also proposed by Myers-Scotton, (Myers-Scotton and Jake 2000). The bilingual Pare and Cushitic pre-Ma'á speakers had no trouble with word order within the noun phrase since the order is Head-Modifier in both languages, Pare and pre-Ma'á. Fur­ thermore they probably had a preference for demonstratives and possessives to be separate words, as is the situation in Pare, and this is also an option in the relevant Cushitic languages as an alternative to use suffixes. Separate words are generally pre­ ferred by speakers who feel insecure about their language. The fact that the possessive pronouns begin in  in Ma'á is an indication for the preference of independent words (pronouns) to suffixes among the Cushitic possessives. The possessive pronouns in several Cushitic languages need a gender prefix ku- (masculine/plural) or ta- (femi­ nine). The Ma'á situation resulted from the generalisation of the more frequent ones, i.e. those beginning in k. This evidence, plus the fact that there is no agreement within the Ma'á NP in possessives, suggests that the speakers lost the Cushitic gender agree­ ment and used only independent forms while generalising the (m/p) forms beginning in k-.24 As Pare was or became their dominant language they followed the Pare catego­ risation of the pronouns, that is to say, the Cushitic m/f sex distinctions in the second and third person were lost.25 The development of the demonstratives was similar. Here too the speakers gave up gender agreement, opted for independent words and followed the Pare categorisa­ tion of a three way distinction. This last aspect probably involved acquiring an addi­ tional form, as Dahalo, Boraana and Eastern Cushitic languages in general have a twoway demonstrative system. In Myers-Scotton's terms, these "early" morphemes sur­ vived, but in a new categorisation, as independent words and without agreement. Genitival constructions were initially formed by juxtaposition and the word final geni­ tive marker got lost through erosion and under the influence of the absence of any

80

CHAPTER 3

comparable inflectional nominal suffixes in Pare. The Pare associative (genitive) be­ tween the head noun and the modifier which agrees in noun class with the head noun was introduced later. Myers-S cotton characterises the Bantu associative as a "bridge" morpheme which is between her "early" and "late outsider". If the influx of Iraqw/Gorwaa-like words correlated with an influx of speakers of Gorwaa, this might have helped the introduction of the associative since the Iraqw-group languages have a similar construction with an anaphoric pronoun following the head noun and agreeing with it. As soon as the associative was introduced, Bantu type agreement entered the Ma'á language. If the speakers were at that time imperfect speakers of Pare then they probably had an impoverished agreement system, both in Pare and in pre-Ma'á. This is suggested by evidence from second language acquisition and from reduction of agreement in Bantu languages in contact; both show that an impoverished Bantu agreement system typically consists of three or four exponents distinguishing human (class 1/2) and non-human (class 9/10) and singular/plural.26 Such a pattern of noun phrase internal agreement is also reported for situations of code-switching. For exam­ ple, if we look at Swahili - English code-switching, we can observe the following agreement patterns within the noun phrase: A Swahili modifier with an English head noun has 1/2 or 9/10 agreement, ((3.23), (3.24)). Foreign (English) modifiers show no agreement ((3.25), (3.26)). An English head noun is either a bare form ((3.23), (3.29)) or it has a plural prefix ma- ((3.27), (3.28)), or an English plural suffix -s (3.29), or both. In all cases the agreement is as stated above; note that class 10 agreement occurs even when the head noun has a class 6 (plural) prefix ((3.27), (3.28)). All these exam­ ples are taken again from Myers-Scotton (1993a). (3.23)yu-le chick y-ako (p. 186, ex. 10) l-DEM3chick 9-your 'this chick of yours' (3.24) our brother w-a Thika (p. 141 ) our brother 1-of Thika 'our brother of Thika' (3.25)mtu innocent (p.29, ex. 8) 1.person innocent 'an innocent person* (3.26) nguo nyi-ngine bright (p.28, ex.7) clothes 10-other bright 'other bright clothes' (3.27)mapeni z-a lunch (p. 171) 6-pennies 10-of lunch

'money for lunch'

(3.28)zi-le matime z-ote z-engine (p. 106, ex. 29) 9-DEM3 6-time 10-all 10-other 'all the other times'

LINGUISTIC HISTORY

(3.29)jokes kama hizo (p.44) jokes like 10:DEM2

'jokes like that'

(3.30)books z-angu(p.80) books 10-my

'my books'

81

The fact that foreign modifiers do not show agreement in code-switching with Swahili as the matrix language shows us how certain modifiers, possessives and demonstra­ tives in the case of Ma'á, can be exempted from agreement within the noun phrase. In fact, non-agreeing modifiers exist both in Normal and Inner Mbugu, and in many other Bantu languages, e.g., in the higher numerals. Other Inner Mbugu modifiers, however, show agreement, in particular in the adjectives. Whereas Normal Mbugu quantifiers have pronominal agreement, their Inner Mbugu counterparts do not; in­ stead they show nominal agreement (see section 7.1). Noun prefixes were added later, once the noun class agreement system was in place. No agreement developed on possessives and demonstratives because the pattern of no agreement for these function words had already stabilised in Ma'á. With grow­ ing knowledge of Mbugu-Pare this rudimentary agreement was later corrected for agreement beyond the NP level and in order to do this, I will argue presently, the con­ struction of a parallel lexicon is a prerequisite. Sentence level agreement developed for possessives. This kind of outside agreement is a later development. It developed only on possessives and not on demonstratives because the Bantu possessives, which by that time formed the dominant model, are combinations of an associative marker plus a possessive stem and the development of the predicative agreement on Ma'á possessives can be seen as parallel to the introduction of the Bantu associative. Before we discuss the crucial final stage of developing a full Bantu noun class agreement, let us hypothesize a scenario for the developments in the verb. Pare, and Bantu in general, has subject agreement, followed by   (Tense-Aspect-Mood) morpheme and optional object agreement preceding the verb stem, while Dahalo and other Cushitic languages have subject and  suffixes.27 Thus the differences here are phenomenal. I can only assume that the Cushitic verb was reinterpreted as uninflected under the influence of the lack (or more precisely, rudimentary status) of verb final inflection in Bantu. Verb final - and polarity- inflection in Bantu (including Pare) does exist but this is typically ignored by speakers with a non-Bantu background who are not fluent in a Bantu language. The Bantu verbal derivations developed at a late stage and they were all additive since there are no semantically and functionally comparable morphemes in Cushitic, with the exception of the causative. The Cushitic causative was retained while the Cushitic middle derivation, which had no parallel in Pare, was lost. The Cushitic imperfective reduplication, which was also without a par­ allel in Pare, was replaced by Bantu TAM-inflection. The passive may have been avoided by using an impersonal construction, which is available in Cushitic languages, especially during the period when Bantu subject agreement on the verb was introduced as a way to avoid using non-human subject agreement on the verb (classes 3-16) of the

82

CHAPTER 3

promoted patient.28 At that stage object agreement, which is not obligatory in Pare, was also avoided. If the pre-Ma'á language in decay had a reduced  marking sys­ tem, the speakers might have introduced Pare  markers as independent words preceding the verb, thus fitting the niche of the Cushitic (e.g. Dahalo, Iraqw) selectors which are  clitics preceding the verb and cliticising to the left, onto the subject pronoun, like the  affixes in Bantu.29 Given the common structural position in Pare and Cushitic of  it is likely that the introduction of Pare  markers into Ma'á speech brought about the introduction of subject agreement at the same time. Again this might have happened in two stages: an initial stage with a rudimentary sin­ gular/plural and human/non-human agreement followed by a later correction to full noun class agreement. Myers-Scotton's studies show that in code-switching with a Bantu language as the matrix language a rudimentary agreement system prevails. Sub­ ject or object agreement with non-animate English nouns in Swahili-English codeswitching is always with class 9 (i) or 10 (zi), as the following examples from MyersScotton (1993a) show: (3.31) results z-ake ha-zi-kuwa nzuri results 10-hisNEG-lO-be good 'His results are not good.' (p. 106, ex. 28) (3.32) ...services amba-zo zi-na-function right now ... that-10:REL 10-PRESENT-function 'services that function right now' (p. 130, ex. 16) (3.33)ni-ka-ona curtain zi-me-fung-w-a lSG-CSEC-see curtains 10-PF-close-PAS-F 'I saw the curtains closed.' (p.95, ex. 22) (3.34)ni-li-i-suggest i-le plan lSG-PST-9-suggest 9-DEM3 plan 'I suggested that plan.' (p. 109, ex. 41b) (3.35)h-a-zi-buy hizo NEG-1-lO-buy 10:DEM2 'He didn't buy beers.' (p.l 15, ex. 51) This last stage of integrating full agreement remains a major step. In order to know the appropriate agreement morpheme for a particular pre-Ma'á noun, whether it is asso­ ciative agreement or subject agreement, the speakers can only resort to the noun class of the equivalent noun in Pare. By that time the meanings of the nouns in pre-Ma'á had already been reshaped according to Pare norms, as often happens in situations of language reduction. Rather, the introduction of, or the expansion into, a full Bantu agreement system has to go hand-in-hand with putting pre-Ma'á nouns on a par with

LINGUISTIC HISTORY

83

Pare nouns, i.e. with the construction of a parallel lexicon. The two processes, the in­ troduction of full Bantu inflection (subject+TAM and associative) and the develop­ ment of a parallel lexicon, cannot be conceived to be separate from each other. Once this Bantu inflection is in place, there is no more grammatical uncertainty with speak­ ers of pre-Ma'á since they can resort to the structures of the language they know bet­ ter, Pare. As a result the use of the language in decay may have gained ground again, assuming that the attitude to the language was always one of an attempt at preserva­ tion, revitalisation or imitation. It is difficult to imagine that the introduction or expan­ sion could have been gradual across the noun classes, say introducing class agreement for class pair 5/6 but not for 7/8, nor are there any indications for such a gradual de­ velopment. The process must have been one major change: the switch to NMb/Pare grammar goes hand in hand to the creation of a full noun class system with non-Bantu lexical material. I would call it language shift.'So far, there is no need to assume de­ liberate or conscious manipulation. But once the parallel lexicon structure is in place, the road is open for lexical extension within this structure. With a growing confidence in the Ma'á lect and access to non-Bantu lexical material the speakers were able to expand their Ma'á vocabulary and at that moment the process became a deliberate at­ tempt at reviving the "old language" and creating their own language for reasons of ethnic identity. The non-Bantu lexical material was available due to an influx of speakers from Gorwaa as well as due to the fact that they were living in a Maasai en­ vironment prior to the settlement in the present area. In conclusion, the development of a parallel lexicon cannot be separated from a shift to NMb/Pare, at least in the sense of a shift to NMb/Pare grammar dominance in the minds of the "bilingual" speakers. However, this does not exclude the possibility or the likelihood of a prior stage of gradual decay of the original Cushitic language coupled with structural influence from Pare. Thus, some initial gradual bantuisation is likely to have happened but eventually there was a rapid process of shift to full Bantu grammar. These last stages that are hy­ pothesised here (switch to Bantu verbal inflection, parallel structures in the lexicon and expansion of the lexicon by lexical manipulation) must have happened before the recorded history of Ma'á (at the latest in the 19th century) since the published materi­ als on Ma'á show all the signs of these processes. Another property that a developing noun class system has to acquire is plural marking by class pairs. The code-switching properties show us that the addition of a noun class prefix on a foreign element to mark plural is common. This is, however, not the regular plural member of a noun class pair, but one particular noun class, namely class 6, that has stronger associations with plural than any other class and is regularly used for collectives in many Bantu languages. A further important considera­ tion is that the resultant form, a foreign element with class 6 prefix, does not have class 6 agreement in code-switching (see (3.27), (3.28)). The Ma'á nouns, however — even the prefixless group mentioned above— take a plural in the regular plural class of the pair. In Ma'á there is no tendency towards having plurals in class 6. Again we need the parallel lexicon to account for this feature of the noun class system.

84

CHAPTER 3

While in Bantu languages number is expressed by means of a plural noun class prefixes of a singular / plural pair, in Cushitic languages number is expressed by suf­ fixes. A small number of Inner Mbugu nouns have suffixes that resemble Cushitic plu­ ral suffixes, in combination with the regular noun class prefixes. Thomason (1997:481) remarks about this that "it seems relatively unlikely that they would have preserved a few relics of Cushitic number marking beside their regular and productive Bantu number marking''30. Thus, the Inner Mbugu suffixes in question were or had become meaningless, having as their only function the marking words as being Inner Mbugu, see section 3.7. The addition of a noun class prefix to a borrowed noun is presumably the last stage. Once sentential and noun phrase agreement have developed, as well as singular - plural pairing, the addition of a noun class prefix is indeed a small step. The higher proportion of class 9 prefixless nouns in Inner Mbugu is due to the fact that the noun prefix for 9 that should be added in paralexification is zero in the speaker's reality be­ cause it has no separable segmental content. The small and stable group of prefixless nouns in classes 3 and 14.1 is a feature of the Bumbuli dialect and shows that paralexi­ fication is possible without the addition of the noun class prefix. This may have been more widespread at an earlier stage. Each of the earlier steps require that the elements of the Ma'á lexicon are equated with Mbugu lexical elements and that the inflectional categories of Mbugu have made their way into Ma'á. This includes the small set of prefixless words in Bumbuli because these are in the appropriate noun class and thus have been put on ą par with their Normal Mbugu counterparts. These two properties already reduce Ma'á to a kind of Mbugu. The present Ma'á noun class system cannot have developed by gradual bantuisation alone, that is, without a shift to Mbugu. The other elements of non-Bantu grammar which are discussed in sections 3.53.9 are not, in my opinion, evidence for gradual bantuisation. For example, I have shown that nhtambala 'snail' is not a Cushitic singulative in -(a)la but a Shambaa loan, next to the Pare nhtamba; that the ending reflecting gender difference between 'brother' and 'sister' is a possessive and has a perfect parallel in Mbugu and Pare; that the words for 'ram' and 'goat' go back to the same root; the apparent different endings are a result of adding a dummy suffix -éno plus the truncation rule and subsequent differentiation of meaning. At any rate Cushitic languages differentiate 'rams' and 'goats' by different lexical roots and not by gender suffixes. I have also shown that the use of the copula, the expression of 'to have', and the colour terms are not Cushitic remnants (see section 3.5). The variation between equivalent forms such as nihi, nihena 'animal', mahlarí, mahlaréno (6) 'clouds', showing endings -ena, -eno, and similar ones suggest a transitional period according to Thomason (1997:481). It is, however, not necessarily a sign of gradual bantuisation; these endings could equally be the remnants of a shift to Bantu. Plural (and other) endings of nouns were consid­ ered part of the stem in Mbugu and when two number forms of a non-Bantu word sur­ vived, these must have been considered to be root variants and not different number forms. This opened the way for these endings to be used as a marker of Inner Mbugu in a situation when such a register was developing.

LINGUISTIC HISTORY

85

The most robust evidence for non-Bantu grammar in Inner Mbugu is the causa­ tive suffix -ti, and similar non-Bantu causative suffixes. I have already discussed this formative in section 3.6. The suffix can be called productive in the sense that it has a grammatical value and that it has been added to a few verbs of Maasai and Pare origin, but the set of verbs in which -ti occurs is closed, and any new causative is made with ij.31 As I argue in section 3.6. this suffix is probably a continuation (both in meaning and in position) of a Cushitic suffix similar to Dahalo id. The mechanism of four part parallelism in which both lexical root and affix are replaced by their parallels (see Hock 1991:171 or Ross 1996:181-183 on metatypy) may have played a role in assign­ ing it a causative meaning (or keeping this meaning alive) and putting it on a par with the productive causative suffix regardless of its origin. Another challenge is the origin of the set of pronouns. Here too the resemblance to Dahalo is greater than to any other possible Cushitic or non-Bantu candidate. In section 3.7 Í have mentioned some other non-Bantu remnants such as -me for abstract entities, an ending -ao, and deviant parallel derivational relations of verbs. If these remnants point to an old Kenyan Cushitic influence, older than the influence of Gorwaa/Iraqw, then possibly this was the older language of the forefathers of the Mbugu. Such remnants of the original language could survive because of the restructuring into parallel lexicons If we approach the history from the other side, i.e. from the present situation, and stretch the explanatory power of lexical manipulation, we could sketch a scenario in which there was no gradual influence from Bantu at all, as we did in Mous (1996). In such a scenario the deviant causative was paralleled to the Pare causative as part of the lexical manipulation, the empty nominal suffixes were borrowed and the demon­ stratives, possessives (and the personal pronouns) were borrowed parallel to their Pare counterparts along with the (basic) vocabulary. This, however, would not explain the deviant agreement patterns within the NP. A period of gradual bantuisation would ex­ plain this latter aspect better and it would also allow for the causative suffix -ti to sur­ vive —be it parallel to the Mbugu equivalent— rather than to be revitalized. The most likely scenario for the linguistic history is one in which a Cushitic speaking people of the "Old Kenyan Cushitic'' type became bilingual in Pare and their original language. Pare became dominant and there was substantial influence from Pare in the speech of their original language. At one point in time, when the Bantu associative and senten­ tial agreement became dominant in the speech of their original language, the original vocabulary was put on a par with the dominant Pare vocabulary. In the context of a strong desire to maintain a language that could show that they were different from their Bantu neighbours, this parallel vocabulary was expanded with non-Bantu mate­ rial, which was available due to an influx of Gorwaa speakers and due to the fact that they were or had recently been in a Maasai speaking environment. The flight to the Usamabara mountains where the dominant environment was Bantu again, strength­ ened the desire for a language of their own. As did the fusion with a different group of Mbugu who came straight from the Pare mountains. When the two groups became one again they went back to South Pare for their initiation. It is not unlikely that the initia-

86

CHAPTER 3

tion period in Vudee which might have included learning a secret language, helped the deliberate expansion of the parallel lexicon. 3.11 Code-switching as a model for Ma 'á Other attempts at explaining Ma'á have been put forward. Prominent among these is the view of Ma"á as "frozen code-switching'". Code-switching is relevant for Ma'á in several ways since it does indeed occur between the Inner Mbugu and Normal Mbugu as well as with Swahili as an embedded language. On another level Ma'á itself has been seen as an instance of code-switching. Thus code-switching could also be taken as a model of how to deal with one grammar and a double lexicon. Finally, it has been proposed that Ma'á developed out of a turnover of the matrix language (Myers-Scotton 1998a). That code-switching does occur is not surprising given the fact that the Mbugu all speak Ma'á, Normal Mbugu, Shambaa and Swahili. Switching between Ma'á and Normal Mbugu, however, cannot be described as normal code-switching in models such as Myers-Scotton's Matrix Language Frame model (Myers-Scotton 1993a), since in these models the switch can be recognised from the use of grammatical features of the other language while in Mbugu there is no difference in grammatical morphemes between the two languages. If we take the matrix language to be Normal Mbugu then every instance of use of an Inner Mbugu lexeme is an Embedded Island and we can no longer recognise when speakers really switch to another language. A similar problem poses itself for switching between dialects that share a com­ mon grammar. Giacalone Ramat (1995:46-47) points out that the crucial factor for the possibility of code-switching is when there is a separate system and not a single sys­ tem with internal variation, but that this distinction is not always easy to make. She argues that code-switching is different from monolingual style shifting (the change of formality levels within the same language) because code-switching is only possible if the two systems are independent and the speakers see them as being different. The Mbugu situation is different from the Italian dialect situation that she discusses in the sense that the structural grammatical differences are absent in Mbugu. Moreover, the Mbugu practice is not comparable to monolingual style shifting even if structurally the comparison does hold. In actual fact, speakers do not speak "more" or "less" Inner or Normal Mbugu but instead they maximise one of the two languages. The two systems are linguistically not independent but in the speakers' views they are separate. While speaking Inner Mbugu they do so as fully as possible; the expected permanent and gradual switch between languages does not occur. When I taped texts, it was surprising how they much the use of one of the two languages was maximised; the texts were usually completely in one and the same lan­ guage; also when several speakers were involved, they stuck to the same language. There are some exceptions to this general picture. In my data collection I have one conversation between four people which is basically in Inner Mbugu; but as two of the participants were clearly more at ease in Normal Mbugu, they show more Normal Mbugu in their speech.

LINGUISTIC HISTORY

87

In other texts, I also have instances of clear and remarkably abrupt and full switches between languages. In the interview presented in section 9.3, which was sup­ posed to be in Normal Mbugu, there is a sharp switch to Inner Mbugu. Both partici­ pants, the interviewer (my research assistant William Mboko (W)) and the interviewed person, Mr Ishika (I), had used one or two Inner Mbugu forms at that point and thus knew that such a shift was feasible. Both speakers were equally comfortable in Inner and Normal Mbugu. The elder, Mr. Ishika, takes the initiative to shift to Inner Mbugu and the interviewer follows immediately and completely despite the fact that he knew that the whole purpose of the taping session was to record Normal Mbugu data. The reason for the shift is presumably the fact that the topic of conversation, the two Mbugu languages, is perceived to be better discussed in Inner Mbugu. Fifteen minutes into the interview they shifted back to Normal Mbugu when a passer-by greeted them in Normal Mbugu. Later the interviewer switched to Inner Mbugu by repeating the previous sentence by Mr. Ishika in Inner Mbugu. All subsequent switches were initiated by Mr. Ishika and related to the topic of conversation. All switches occured at sentence boundaries. The passage in section 9.3 shows that a shift between Inner and Normal Mbugu is possible and easily recognised because such a shift is from one extreme to the other. The reasons for shifting are those that are com­ mon in unmarked code-switching: topic related and change in participants in the con­ versation in accordance with the Unmarked Choice Maxim proposed by MyersScotton (1993b and 1998b). In fact I have no instance of a Marked Choice in InnerNormal Mbugu code-switching in my data and the reason is that the possibility of its use and not the use itself marks Inner Mbugu ethnicity. People who only speak Nor­ mal Mbugu are not considered less Mbugu and thus the need to distance oneself from Normal Mbugu by using Inner Mbugu does not arise. Myers-Scotton's exploratory choice maxim (checking out competence and preference) and the virtuosity maxim (adapting to all speech partners) play a role but not the deference maxim (when re­ spect is due) which does play a role in comparable parallel lexicon situations such as registers of respect (hlonipha), (Myers-Scotton 1998b). 3.12 Mixed languages as a challenge to historical linguistics Mixed languages have in common with pidgins and creoles that their existence poses a challenge to historical linguistics and to language classification. The term "mixed language" has a bad name for those who consider it a denial of the laws of historical linguistics (cf. Van Driem 2001:163-164, Kortlandt 2001:123). In the field of African linguistics, Wolff (1965) and Heine (1969), among others, have warned against the use of the term. Wolff criticises Mukarovsky's characterisation of Fulfulde as "Mischsprache" with a Niger-Congo substrate and a superstrate that is called "Mauretanisch". Wolff complains about a lack of rigor in recognising sub- and super­ state influence and in drawing conclusions from such influences. He also points out the lack of evidence of how a population can haphazardly select subsystems. Heine shows that Meinhof, following Lepsius, used the term "mixed language" in a very loose way, that is, as the result of any contact between unrelated languages. On the

88

CHAPTER 3

basis of his own fieldwork and research on Mbugu, Heine argues that the classification of Ma'á depends on the question whether preference is given to lexicon or grammar and he concludes that Mbugu is indeed a mixed language with respect to its genetic classification comparable to Armenian Romani as described by Meillet (1914: 420).32 Greenberg, however, does not consider Ma'á to be the exception that proves the rule. He uses the title "Are there mixed languages?" for his discussion of Thomason and Kaufman's proposals in particular with regard to Ma'á and Mednyj Aleut (Greenberg 1999). He remarks that "the problem with the hypotheses of Ma"á as a mixed language is that those who espouse it have treated it as a synchronic problem" (Greenberg 1999:628). Thus there are two problems with the term "mixed language". One is that it is has been used without restraint for all kind of mixtures, and the second is that if it is used in a diachronic sense it challenges the overall validity of the comparative method, see also Bakker and Mous (1994: 4-7). I agree with Greenberg that the term is a syn­ chronic characterisation of a mismatch between the grammatical and lexical elements and that it does not reveal much about the question at stake, that of the diachronic de­ velopment of the language.33 The mixed nature of Ma'á is primarily the result of the diachronic process of parallel lexicon building. Ma'á (Inner Mbugu) has often been quoted as the exception to the rule that mixed languages do not exist. One could argue with Van Driem (2001:165) that by reducing Ma'á to Inner Mbugu, to a register-like phenomenon, it no longer poses a challenge to historical linguistics, because it remains possible to distinguish between inherited and borrowed components. Related to the issue of the definition of a mixed language, we are faced with the problem of the classification of such languages. Brenzinger (1987) gives a historical overview of the various opinions on the case of Ma'á, which has been considered as Cushitic, as Bantu, and as unclassifiable. Similar views on the history of Ma'á can even result in opposite stands on its classification. Thomason (1983) and Thomason and Kaufman (1988) argue against Greenberg (1966), and Greenberg (1999) argues against Thomason and Kaufman (1988), yet both consider Ma'á to be a separate lan­ guage evolved from a Cushitic language and both assign a crucial role to the fact that Ma'á's basic vocabulary is Cushitic. Greenberg's main argument is that the numerals 1-6 and 10 are non-Bantu and the numerals 1 and 2 are Cushitic. "It is a well-attested phenomenon that when a language borrows numerals from a foreign source, it is the lowest and most unmarked numerals which will be of indigenous origin as well as the numeral for the base of the system, most commonly ten" Greenberg (1999:630). Our research confirms roughly that state of affairs in Mbugu numerals, see Table 3.12. The Normal Mbugu numerals are all identical to Pare. In reaction to Greenberg I could argue that an observed statistical universal cannot serve to determine the direction of borrowing. But a more serious problem seems to me that Greenberg's tacit assumption is that the numerals in question are from one source. In my analysis this is not neces­ sarily the case, and there is no positive indication that they are indeed from one source. Greenberg himself makes already a division that some numerals are Cushitic while others are non-Bantu of unclear origin. In fact there is an indication that the numbers 'two' and 'ten', on the one hand, and the number 'three', and the number

LINGUISTIC HISTORY

89

'four', on the other, have different sources. Some of these basic numbers are from Bantu, number 'one' is from Pare (ultimately possibly from Cushitic) and the alterna­ tive word for 'tens* in Inner Mbugu, mi-rongo (4), is also from Bantu, cf. Davidą miron go. # 1 2 3

NMb -mwé -iri -tátu

1Mb -wé ' -nú -kái

4 -nne -hái 5 -sáno -kó i 6 m-tandát(u) mátez  1 (m)-fúngáte - =NMb = NMb 8 m- = NMb 9 ken dá i-kádo, 10 i- i m-kadéma

source for Inner Mbugu Normal Mbugu unclear origin possibly from Maasai okuni '3' (cf. West-Rift Southern Cushitic tam, Proto Eastern Cushitic sed(d)ah) Qwadza/Aasáx haka Cushitic, cf. Gorwaa kooán, Proto Eastern Cushitic kenunclear origin from Pare from Pare from Pare possibly related to Oromo: kud'aani plus vowel metathesis.

Table 3.12: Inner Mbugu Numerals and their origin The centrality of classification in the discussion about the validity of the comparative method is problematic. After all, classification is a by-product and not the essence of the comparative method. Historical linguistics is about understanding language change, not about classification. Indeed if classification were a goal in itself it should be hierarchical, i.e. excluding double membership, and all-encompassing, that is, clas­ sifying every single language, and cases such as Ma'á would be highly problematic. As argued by Owens (1991), if the comparative method encounters cases whose his­ tory can be understood but that pose difficulties with classification, this is not a prob­ lem. To me the question of the classification of Ma'á can be compared to that of whether to refer to a transvestite as "he" or "she". At first sight Ma'á may seem to be Cushitic; closer inspection reveals that it is Bantu but trying desperately to hide that fact. Once we know the reality we may still feel uncomfortable with calling Ma'á Bantu when the core vocabulary is clearly not Bantu but that does not mean Ma'á is unclassifiable: It is a Bantu language even if the speakers want it to be non-Bantu and even if the forefathers spoke a Cushitic language. Brenzinger (1987) and Sasse (1993) have argued that Ma'á is the result of an at­ tempt to undo a shift to Pare when it was "too late" (U-turn hypothesis). In their analysis, the speakers tried to (re-)learn the language that they had already given up. Creating a lexical register is certainly a way to do this, although, even if the intention may be to approximate the "old" language, the result is clearly something quite differ­ ent. Sasse observes that language death situations may give rise to special languages that form a hybrid with the morpho-syntactic matrix coming from the target language and the vocabulary being a residue of the abandoned language. He mentions Kre-

90

CHAPTER 3

konika, a secret language with Greek morpho-syntax and most of the secret vocabu­ lary from Arvanitika (Albanian in Greece), and he points to Ma'á. He further remarks that "no plausible theory' of their [such hybrid languages] origin is as yet in sight ..." (Sasse 1992:23, and footnote 16). He suggests for Ma'á that this "hybridisation" is a relexification of the target language after the shift had already taken place and when vocabulary of the abandoned language was still present. This vocabulary was inserted into the morpho-syntactic frame of the target language. The perspective on Ma'á that I give in this work makes it a valid case in point for Sasse's suggestions. In addition, it offers a contribution to such "a theory of their origin" and shows in what form the hy­ pothetical "relexification" took place. Accepting that Ma'á is not a threat to the comparative method does not mean that the case of Ma'á has little to offer to historical linguistics. The creation of the par­ allel lexicon, of course, involves heavy borrowing but it is borrowing of a very special kind. There are a number of properties that make it unusual: 1. it is both additive and replacive at the same time; it is replacive in the sense that words are borrowed for concepts that already exist and it is additive in the sense that the original term is not lost but continues to exist, on a par with the new term; 2. basic vocabulary is affected heavier than marginal vocabulary; 3. words may be deformed by special processes. These properties apply not only to the Ma'á case but to all cases of lexical manipula­ tion and parallel lexicon building in general (see Mous 2001b, to appear). One ap­ proach to such lexical manipulation is to exclude these cases from the comparative method. Dixon (1997) argues that cases like Ma'á defy the comparative method and that areas where lexical manipulation played an important role over long periods of time, cannot be tackled with the family tree model and the comparative method. It is my hope that more research into lexical manipulation will show that it has laws of its own and that the comparative method may be extended with a limited machinery for the lexical manipulation cases. One such law is that formal (non-semantic) morphological compartimentalisations of lexicon, such as a noun class system, are obeyed by lexical manipulation. Non-related forms that agree in noun class would point to lexical manipulation. This does not help of course for language families where there are no noun classes, and the argument is already questionable for gender systems with only two or three exponents. Still, it is an interesting line of research to investigate paralexification as a diachronic process and specifically in the history of other mixed languages. In Michif, for exam­ ple, nouns are typically from French while the rest of the lexicon and the grammar is from Cree/Ojibwa, see Bakker (1997). However, in addition to French grammatical nominal features, nouns have a formal "animate" / "inanimate" distinction that paral­ lels the same distinction in Cree, see Bakker (1992:93); thus largent 'money' is in­ animate, parallel to the Cree word for 'money'. This suggests that at some stage, when speakers were still fluent in Cree, they equated French nouns with Cree nouns; in other words, Michif originated as a parallel lexicon whose basic nominal Cree/Ojibwa component was subsequently lost.34 The process of paralexification can still be wit­ nessed in arbitrary morpho-lexical properties such as the "animate" / "inanimate" dis-

LINGUISTIC HISTORY

91

tinction, since these are shared in parallel lexicons as a consequence of the fact that the borrowing is both additive and replacive at the same time. A second lesson to be learnt for historical linguistics is that as soon as lexical manipulation operates, core vocabulary is the area most affected. This is already clear from the fact that the parallel borrowing is additive and the conservative nature of core vocabulary in general is based on the concept of resistance to loss; there is no loss in parallel borrowing. In addition, lexical manipulation is most effective precisely in the core vocabulary when its function is to express identity or to be secretive. Lexical manipulation complements the comparative method, but it also raises the question whether all mixed languages arose through parallel lexicon building along the lines I suggest for Ma'á. An alternative process that could lead to a mixed language of the Ma'á type would be fossilisation of code-switching. Mazrui (1995) proposed this as an important process in the development of Sheng, a code-switching variety of Swahili that is used by youth in Nairobi. Certain lexical items in Sheng are always taken from the embedded language (mostly English); it is part of the characterstics of Sheng that these key words for the youth are not taken from the matrix lan­ guage (Swahili) but from the fashionable language, English (or Luo). Although the degree of mixture of Sheng does not come close to that of a language such as Ma'á, it is conceivable that the processes that are at work in Sheng result in structures that are comparable to mixed languages such as Ma'á. The Matrix Language turnover hypothesis as advocated by Myers-Scotton (1993a:220-223, 2002) attributes a crucial role to code-switching in the history of Ma'á. In her view, Ma'á is the fossilisation of a late stage of turnover of the matrix language to Mbugu. In Myers-Scotton (1998: 306-310, 2002) she discusses her and my analysis of Ma'á. The position I take in the present work is slightly different though from that in my earlier publications. Both Thomason's (1997) critique and Myers-Scotton's development of her Code-Switching model introducing a distinction between "early" and "late" system morphemes have persuaded me to accept the likeli­ hood of survival of structural features of the original language. In her latest work (2002) Myers-Scotton proposes seven stages in the history of Ma'á. She proposes that in her stage four extensive code-switching between original Ma'á and Pare led to con­ vergence to Pare at the levels of abstract lexical structure. I assume that her stage five, dominance of Normal Mbugu, or some form of Pare, was contemporary with her stage four. I agree that the bilingualism in dominant Pare reshaped the lexical structures of their original language. In her stages six and seven the abstract grammatical frame is gradually modified introducing surface-level Bantu system morphemes and subse­ quent gradual bantuisation of late system morphemes and finally the entire grammati­ cal system became identical to that of Normal Mbugu. I have already argued why I find it highly unlikely that this happened gradually. Besides the structural difficulties that a gradual adaptation would entail, one also has to bear in mind that in MyersScotton's scenario the speakers have only two varieties at that moment: Pare, the dominant language that they use for outside communication and fossilised codeswitching (between their original language and Ma'á) for in-group talk. This means

92

CHAPTER 3

that they had already lost their original language and that Pare grammar was their dominant grammar. In my scenario this shift to Normal Mbugu grammar in the "out­ side late system morphemes", i.e. to full noun class agreement at sentence level, did not happen gradually.35 Moreover, this shift is a pre-requisite for the parallel lexicon building that I propose as the crucial process in the history of Ma'á. Myers-Scotton concludes her scenario with the remark that the lexicon remained largely intact. In my analysis, the lexicon is everything but intact and in fact only a small portion of the Ma'á lexicon can be attributed to the hypothesized earlier language; other sources be­ ing Gorwaa, Maasai and manipulated Normal Mbugu. The major difference of opinion between my view and hers lies in the fact that she does not accept the parallel lexicon building on principal grounds. In my view it is the same principle that grammar is automatic, that is, not available for conscious manipulation, that is behind both the structure of code-switching and the fact that linguistic manipulation is restricted to the lexicon. Another reason why Myers-Scotton's story is incomplete is that Ma'á speech cannot have been fossilisation tout court because no code-switching takes place by rule at every lexical root the way Ma'á does. Ma'á and other mixed languages have led to proposals of socio-historical mod­ els of how they could arise. Two proposals are of interest to us: Bakker's generalisa­ tions on the emergence of mixed languages, and Thomason's model. Other proposals for linking types of language change with sociolinguistic variables (kind of emigra­ tion, linguistic make-up of area of settlement, marriage customs), such as those found in Guy (1990), Ross (1991) and Reh (1995) do not include mixed languages. Bakker (1997:204-208) found that there are two types of social situations that typically result in mixed languages. One is the situation in which a community follows a pattern of parallel mixed marriages whereby all mothers speak one language and all fathers an­ other one and the children try to approximate the father's language by replacing lexi­ cal material from their mother's language with material from their father's language. He mentions Michif, Island Carib, Basters and Griekwas, Peco' and Krojo as exam­ ples. The other type of situation (nomads' languages) is one of a nomadic group with a professional specialisation that is viewed as belonging to a low-caste by the societies in which they operate. The language of the dominant society furnishes the grammar but the original language survives in the lexicon. Examples are the Para-Romani varie­ ties. In both situations identity or ethnicity plays an important role and in the second one possibly also secrecy. In the first group the language is the result of a strong posi­ tive attitude towards the identity of the fathers or of a new society; the embedded or parallel lexicon is derived from the language associated with the fathers only. In the second group the attitude is more defensive against the dominant surrounding society and the embedded or parallel lexicon is the original language but may be enriched by other deviant material, as Matras shows for Jenisch used by commercial nomads in south-west Germany (Matras 2000:86). There is no strong evidence that Ma'á falls into any of these two groups; there is weak evidence that a group of Gorwaa men may have married Mbugu women, but the identity attitude is defensive against the Bantu dominance rather than an attempt at achieving Cushitic identity.

LINGUISTIC HISTORY

93

Thomason (1995) also distinguishes two types of mixed languages: slowdeveloping mixed languages that are the result of a long development of languages under siege; the other type contains mixed languages of new social groups that have arisen in a short period of time. The first group can be compared to dying languages in the sense that they slowly and gradually lose the former grammatical system, and lin­ guistically these languages can be recognised by the fact that the influence of the re­ placing language can be seen in all areas of the language. Thomason argues that Ma'á falls into this first group of mixed languages. Although I fully agree with Thomason's point of departure that mixed languages, extraordinary as they may be, are the result of ordinary processes of language change and I also agree that not all mixed languages form a single set historically (Thomason 1995: 15,17), as will be clear by now, I have my reservations of putting Ma'á in the group of slow adapters.

4

PHONOLOGY

In this chapter I first introduce the segmental units such as consonants and vowels and their phonetic realisation as well as allophonic variation. The inventory is then com­ pleted by a short paragraph on tone. Following the introduction of the segmental and suprasegmental units, the syllable structure and the structure of the phonological word are presented and the phonological and morphophonological processes are discussed. The differences between Inner Mbugu (Ma'á) and Normal Mbugu lie mainly in the fact that Inner Mbugu has a number of additional consonants that are not found in Normal Mbugu, namely, the glottal stop, the lateral fricative, the velar fricative and the prenasalised velar fricative. Other differences are mentioned where relevant, but in the final analysis the phonological differences are not substantial and do not prevent a presentation as one phonology. In this and subsequent chapters I discuss Inner Mbugu and Normal Mbugu simultaneously while pointing out any differences that exist be­ tween the two languages. 4.1

Consonants The consonantal units in a broad phonetic approach are presented in Table 4.1 using the orthography adopted in this work. The consonants in italics are those which only occur in Inner Mbugu. The deviations from the phonetic alphabet are the follow­ ing: ■ hi = [ł], a voiceless lateral fricative, ■ ch = [c], a voiceless palatal stop, ■ j = [f]. a voiced palatal stop, ■ sh = [δ], a voiceless palatal fricative, ■  = [j]a voiced palatal glide, ■ ny = [ŋ]a palatal nasal stop, ■ x = [x], a voiceless velar fricative, ■ gh = [y], a voiced velar fricative, ■ ' = [?], a glottal stop, ■ mb = [mb], a voiced prenasalised labial stop, ■ nd = [nd], a voiced prenasalised alveolar stop, ■ ng = [Dg], a voiced prenasalised velar stop, ■ mhp = [mp], a voiceless prenasalised labial stop, ■ nht = [ft], a voiceless prenasalised alveolar stop, ■ nhk = [ŋk], a voiceless prenasalised velar stop, ■ nhx = [ŋx],a voiceless prenasalised velar fricative.

96

CHAPTER 4

labial P b f v

w m mb mhp

alveolar t d s z 1 r M n nd nht

palatai k ch j sh

velar

glottal

g X

h

gh

 ny

Į)

ng nhk nhx

Table 4.1: Mbugu consonants One of the organising principles of the set of consonants is the opposition voiced voiceless. All the voiced stops are optionally realised with larynx lowering (implosives). This is most common and strongest (or perceptually easiest to recognise) for the labial stop b [6]. Implosives are never in contrast to pulmonic stops. The implosive pronunciation of voiced stops is not peculiar to Mbugu but is, in fact, a feature com­ mon to all the languages spoken in the coastal areas of East Africa, including Swahili. Another basic opposition is that between stops and continuants. Some of the (morpho)phonological processes apply to all the continuants. Table 4.1 distinguishes five places of articulation taking bilabial and labiodental articulation in one column. The labiodental fricatives ƒ and v have the teeth quite far on the lower lip. The palatal consonants are of a different order. One structural peculi­ arity is that they have no prenasalised counterparts.1 Note that the ch is rare in Inner Mbugu. It only occurs in the words churú 'bull' and ku-chi'i 'to carry', and in the lat­ ter the ch is generally replaced by t in the Magamba area. The glottal consonants are special in various respects. The glottal stop is limited to words in Inner Mbugu.2 The glottal stop and the approximant h are not stable in that they sometimes alternate with their absence, that is, the glottal stop can be realised as hiatus in kubú'u or kubú-u 'to enter', mi'azé or miazé 'days'. The lateral voiceless fricative hl is limited to Inner Mbugu. It is an uncommon sound in the area. The sound occurs more often in Southern Cushitic languages and does not generally occur in Bantu languages outside of South Africa. The Taita Bantu language Sagala, however, must have had it at an earlier stage (see section 2.7). His­ torically, there is some oscillation between hl and s (see section 3.1). Occasionally there is variation between the lateral approximant / and the alveo­ lar trill r, in, e.g., tulevona or turevona 'we saw' (NMb), and also in the various forms of the applied extension (see section 5.6). Despite this variation, / and r are clearly distinct phonemes as the (near-)minimal pairs in Table 4.2 show.

97

PHONOLOGY

form with /

hàla gali ikóló kalá kelia (m) kolé

meaning dry try, touch hen bitter; hunt pass kind of room

form with r hará igári ikoré kará kéria korė

meaning calf car log front rat pot

Table 4.2: (Near-)minimal pairs of\ and r Glides occur word-initially and after consonants both as a result of coalescence and in non-derived stems. Word-finally the glide y can only occur after a as in My 'three' (1Mb), which forms a minimal pair with Mi 'push' (1Mb). Word-initial glidés are found in: wé 'buy' [1Mb], wé 'you' [NMb], waha 'drink' [1Mb], ya 'this' [1Mb], yó 'today' [NMb]); word-medially they are found in, e.g., ka-wahá 'knife' [1Mb], luháwa 'ladle' [1Mb], i-wawá 'wing' [I+NMb], ka-zewé 'secretary bird' [I+NMb],- jéwá 'white' [NMb], i-wíwí 'weeded grass' [I+NMb], ayà 'grandmother' [1Mb], i-hóyu 'axe' [1Mb], lu-zoyå 'feather' [I+NMb], báyu 'red' [1Mb]; glides after consonants oc­ cur in: gwá 'steal' [1Mb], gwira 'seize' [NMb], bwa 'side' [1Mb], bwange 'male' [NMb]. 4.2

Spirantisation in velars Some of the sounds given in the table of consonants, Table 4.1 above, are at the same time the basic realisation of one phoneme and an allophone of another phoneme. This is in particular the problem with the velar obstruents. The voiced velar stop and its fricative counterpart are in opposition, thus phonemically distinct, but at the same time the stop can optionally be realised as a fricative in some but not all words. In the voiceless domain a similar spirantisation phenomenon is in operation whereby  is realised as x and x as h while all three are also phonemic. This spirantisation may be regarded as a sort of sound change in operation, spreading faster in one geographical area than in others, occuring in some words but not in all and being more common among younger speakers. The spirantisation of the voiced velar occurs in South Pare too and is wide­ spread in the Rangwi area which is most influenced by the presence of Pare speakers. Variation between g and gh has been recorded in many words, only some of which are presented here: gujú or ghujú [1Mb] 'to swallow', magamba or maghamba [I+NMb] a place name, mgeni or mgheni [NMb] 'visitor', mgahló or mghahló [1Mb] 'visitor', hagia or haghia [NMb] 'sweep', igomé or ighomé [NMb] 'trunk', igana or ighana [I+NMb] 'hundred. This spirantisation phenomenon is also observable in recent loanwords from Swahili where g is generally realised as gh: mbogha 'vegetables' from Sw mboga, mbogho 'buffalo' from Sw mbogo, mghóngó 'back' from Sw mgongo.

98

CHAPTER 4

The two sounds g and gh are in phonemic opposition as can be seen in the (near-)minimal pairs in Table 4.3. Moreover, a distinction between the two sounds also has to be made intervocalically: ighondá 'loud voice' (I+NMb) but igono 'gonor­ rhoea' (I+NMb), magalé 'maize'-(1Mb) but magheri 'evening' (I+NMb), ógha 'be­ witch' (NMb) but ogoha 'fear' (I+NMb). The spirantisation only applies to certain words. For many words the g cannot be realised as a fricative and vice-versa. Words containing g that do not have a variant with gh are: ga 'die' (1Mb), ge 'open' (1Mb), géra 'bring' (1Mb), gera 'put into' (NMb), gugulu 'run' (1Mb), go 'suck' (1Mb), gomae 'clothes' (1Mb), gurà 'fetch' (NMb), igeru 'banana' (1Mb), igi 'egg' (1Mb), kigi 'thing' (1Mb), lugo 'fence' (NMb) (lughó was considered Pare), and others. form with g gó gera gusha

meaning suck put into give birth

language 1Mb NMb NMb

form with gh ghó gherá ghusha

meaning my metal explain

language 1Mb NMb NMb

Table 4.3: (Near-) minimal pairs of g and gh Most instances of g are word initial or at least stem-initial, but certainly not all: ichagá 'clitoris' [I+NMb], izogéra 'palm of hand' [1Mb], kihága 'heel' [I+NMb], kizóga 'shoulder' [1Mb], kuga'á 'white, bright' [1Mb], mkálágázo 'Saturday' [I+NMb], kigógo 'bare stalk of banana' [I+NMb], mlogojo 'follower', togola 'thank, praise' [I+NMb], luhige 'door' [1Mb], mlagé 'woman, wife' [1Mb], kiharige 'tail' [1Mb], izigizigi 'mane, neckhair of donkey or goat' [I+NMb], mzigi 'medicine' [NMb], agirú 'male' [1Mb], zógu 'look after' [1Mb], igii or egu vagii 'his brother/father' [1Mb], gugúlu 'run' [1Mb], ngwégu 'grasshopper sp.' [I+NMb], ogu 'castrated goat' [1Mb], nhtwiga 'giraffe' [I+NMb], sígala 'cigarette' [I+NMb]. The only real context restrictions are that gh does not occur in a nasal+obstruent combination, i.e. not ngh, only ng.3 Nor does it occur before w. Note that nhx does oc­ cur, be it sporadically, and thus fricatives are not totally excluded in such na­ sal+obstruent combinations (see section 4.3). The other context where gh is excluded is before a glide w. Nearly every consonant can be followed by w, but not gh, with the exception of passive verb forms such as jughwa 'be opened, be allowed to marry' (NMb). The same variation is present in Pare. Pare has gh where Normal Mbugu has gh or allows gh, as in kighera 'iron', ghuha 'marry', mgheni 'stranger', haghia 'weep', ighana 'hundred'4 . Kotz (1909) describes South Pare as only having g and does not mention gh. Hohenberger (1929/30) working in Shighatini in North Pare reports ex­ tensive use of gh. Kähler-Meyer (1962) describes the North Pare variety also with gh, which is in line with Odden's assertion: "The voiced velar (not preceded by a nasal) in SK [South Pare] is generally the stop g (alternating rarely with y, whereas in NK [North Pare] the voiced velar is always a fricative" (Odden 1986: 263). However, Ka-

PHONOLOGY

99

gaya (1989) has g and gh without explicit mention of variation for the Mbaga (and Gonja) dialect of South Pare. Gweno is like North Pare in having predominantly gh; the latter possibly being influenced by the former, see Mreta (1998), Philippson and Nurse (2000). Shambaa has both g and gh and no alternation, see Roehl (1911:7), LangHeinrich (1921), Kähler-Meyer (1962), Besha (1989:45), while Shambaa's clos­ est relatives, Zigua and Bondei, have g. It should be noted that Shambaa has gh while South Pare, which is otherwise much closer to Shambaa compared to North Pare, seems to have been only recently affected. A further piece'in the puzzle is the fact that dialectal variation within Mbugu shows a greater use of gh for g in the Rangwi area. This area is mostly affected by intrusion from South Pare. Thus the development of g > gh can be attributed to influence from the linguistic environment, South Pare and Shambaa. Nurse (1979:108) argues that the weakening of intervocalic g "tends to affect individual languages rather than whole groups: it replaces [g] in part of Thagicu (Ki­ kuyu), parts of E[astern] Nyanza, part of Seuta (Shambála), part of the Ruflji group (Ngindo), parts of the S[outhern] Highlands, Lugulu, and part of W [est] Tanzania (Nyaturu). It also affects Pare, Saghala, Dawida and Chaga." Thus, the weakening of g is a widespread phenomenon in Bantu languages of the area and at the same time it is not very diagnostic for identifying influences in present-day Mbugu. The Bantu nature of the sound gh can still be seen in the distribution. The sound gh is more common in the Normal Mbugu lexicon: 8.58 % of the dictionary entries as opposed to 3.21 % in Inner Mbugu. Among those items within 1Mb that contain gh only three can be traced back to a non-Bantu source: ghujú or gnjú [1Mb] 'swallow' cognate with Iraqw guu? vi'aghu [1Mb] 'food' cognate with Iraqw Tagmu 'food',5 and mharégha 'arm' cognate with Oromo harka.6 Sagala (Bantu) buga 'affliction, trouble' became bughú (also bughéno) 'fear'. These examples show that the rule operated in the latest Mbugu period. A confusing case is Inner Mbugu gugalu 'run' for which Meinhof (1906) has ghnghulu 'walk away' and which could be from Oromo gugud'd'a but it is also present in Shambaa ghuluka, and in Zigua guluka. A number of instances of gh are, however, clearly the result of recent influence from Shambaa. Shambaa has gh word-initially where Pare has k, e.g., Shambaa ghubika 'to cover' vs. Pare kúbwíka 'turn down, cover oneself, and Mbugu has the Shambaa sound, cf. ghubika 'to cover, close' [NMb]; this correspondence  - gh is used in the distribution over Inner and Normal Mbugu of the word 'mountain, rock': ighámba (5/6) [1Mb] from Shambaa ghamba versus ikamba [NMb] from Pare ikamba. The spirantisation seems to be expanding to affect the voiceless velar resulting in a voiced fricative in the variation between [NMb] nhkughu and nkuku (9/10) 'chicken' from Pare nkúkú, cf. Shambaa nguku. Spirantisation of the voiceless velar without change in voice is very common in the Inner Mbugu part of the lexicon and this is dis­ cussed in the next paragraph. In Inner Mbugu, but not in Normal Mbugu, there is variation between , x and h\ mainly dialectal variation between  and x and between x and h. Yet the three sounds are in opposition as can be seen from the invariant Inner Mbugu words in

100

CHAPTER 4

Table 4.4. This variation  - x - h is limited to Inner Mbugu as is the sound x\ the op­ position  - h is unproblematic in Normal Mbugu. Form with  meaning bury ki kiríi return put high ka'a kuru cultivate

form with x Xi

xiri xá'a xúru xehlú

meaning scrape mix split, tear' pull rat sp.

long ago

kaé

form with h hi hiri (= hà'a) (= h u r u héhla haé

meaning sew, mould hurt

meet far away

Table 4.4: (Near-) minimal pairs  - x - h in Inner Mbugu Some evidence of the variation is given in Table 4.5. The place names Magamba and Rangwi refer to where a specific form was recorded and this does not necessarily rep­ resent systematic dialect variation. Bumbuli 'áxu ixádo m kurá mxoxéra vuxúso ka xwá zóxo mtataxó xabúna áxo ihó ixutáo kixojá luxemú lux arému mxatú shuxé táxo xókwa xúru

Magamba 'áku ikádo mxorá mkokéra vukúso ká kwá zóko mtataxó / mtatakó kabúna xaxa / kaka áho ihora ihuréa ihutáo kihojá laháo luximú, pl: himú luharému mhatú shuhé táho ximéno hókwa húru

Rangwi ihádo mkorá

xa

mtatakó xabún

ixóra ixuréa

laxáo luximú luharemú

himéno

meaning shine, be white ten doctor, circumciser throat, dewlap bow get up be, become sit, reside trap all divide see pit crow hoe market bile finger-nail horn tree, afterbirth leopard pass bird be met, set (of sun) pull

101

PHONOLOGY

Bumbuli xwá'a xwa'ú mxá

Magamba hwá'a mhá xá'a / há'a xála / hàla

Rangwi hwa'ú mxá

meaning return last night medicine split, tear dry

hàla Table 4.5: (Dialect) variation between  - x - h in Inner Mbugu My dictionary files contain only eleven lexical entries with a velar fricative x that is not interchangeable with h or k, two of which are part of a minimal triplet (above). For some of them I only have information from one of the dialect areas or only one in­ stance of the word. These eleven items are: xwahlé 'partridge', xiri 'mix', xi 'scrape', xehlú 'rat sp.', dadáxa 'hang up' (Bumbuli), i'axi 'boil on skin' (Bumbuli), kanyaxe 'baby' (Rangwi), kituxú 'boil on skin', xái 'open the mouth', xara 'itch' (Bumbuli), xumutu 'buy (?)'. This means that the functional load of the voiceless velar fricative is extremely limited. The spirantization of the voiceless velar stop was possibly already present in the (Eastern-)Cushitic language that the Mbugu once spoke; it is present in contemporary Cushitic languages, in particular in Boraana Oromo which has both k→ x and k' → x' and unpredictable allophonic variation (see Stroomer (1995:11) and Owens (1980: 143)). The development k> x> h can be illustrated with the word for 'hoe' which was hikutao in Farler (1885), xutao in Meinhof (1906), ixutáa, ixutáo, in Tucker and Bryan (1974) and ihutáo in present-day Inner Mbugu. Not all historical evidence shows such a unidirectional development; for example, the word for 'ten' develops from h to x to x-k in the documented history of the language: muhadema (Farler 1885), hadu (Storch 1895), xado, mixado, or mixadema (Meinhof 1906), makadu, maxadu or maxademu (Green 1958), ixádu (Tucker and Bryan 1974), and present-day Inner Mbugu has ikádo, ixádo, or mkadéma. A similar "return" to the stop can be seen in the word for 'blood' which is msako 'body' in Farler (1885) saxo in Meinhof (1906) but in presentday Inner Mbugu it is sako as it was in its source sakame (NMb and Pare). There is ample historical evidence for the development x > h see Table 4.6. However, the "res­ toration" of the stop from the fricative is equally well attested historically. In Table 4.6 and Table 4.7, M06 refers to Meinhof (1906), C33 to Copland (1933/34), G58 to Green (1958), and T&B74 to Tucker and Bryan (1974). This is evidence for a con­ tinuation of variation through the last century. meaning swell shoe bird chew tree

form yaxi kizixu ximeno saxo mxatu or mxato

attestation M06 M06 M06 M06 M06

present-day yáhi, yáxi kisuhéra himeno, ximéno saho mhatú, mxatú (Bum)

102

CHAPTER 4

attestation form M06 xai C33,G58 taxo C33 axo C33 vaxe C33 axoni C33 xo C33 texelo G58 mbaxu T&B74 kixoja Table 4.6: x>h in history

meaning yawn pass see people what go there is no sibling market

meaning three egg return pot side-dish bow boil

form xai exoxoha xeriti xore mbux

attestation M06 M06 M06 M06,G58 M06

VUXUSU,

vuxuso to toxo

M06 M06

give birth neck ill person

waxa muluxa

slim all heart

xita xabuna swaxo

spoon

lumexo

2SG

xé

hlxa M06 M06 M06, T&B74 C33 C33 T&B74 T&B74

present-day hài, xái táho áho vahé ahoni ho'o tehélo mbahu kihojá, kixoja (Bum)

present-day kai ikokohá kiríti, kiríi koré, xoré (Bum) mb uka vukusó, vuxusó (Bum) totóko, totóxo (Ra) hlaka waka mrukáo

source Ir: qanhhi Ir: kif OR: okotee N Mb/Pa: mbogha 9

Pa: toghotia, Ma: itokitok Ir: wakri 'chin'

kita kabuna swaho (Bryan: swako) luméko (Bryan: lomeko) T&B74 ké Table 4.7: "Restoration " of x > 

The spirantization of the voiced velar stop is a phenomenon that is typical for the Bantu languages of the area. However, the spirantization of the voiceless stop is typi­ cal for the Cushitic languages of the area. Ma'á shows both phenomena.

PHONOLOGY

103

Prenasalised obstruents The prenasalised stops are considered to be separate entities and not a sequence of phonemes. Taking them as single units simplifies the description of the syllable structure (see section 4.6). Moreover, an analysis as a sequence of phonemes would involve postulating a series of voiceless nasals that only occur in these sequences (or a devoicing rule). Nasal-stop sequences do occur but only with syllabic nasals. In this case, one remarks a clear difference in duration of the nasal onset: The prenasalised stops have a short nasal onset; in the nasal-stop sequences the nasal part is syllabic. The place of articulation does not change within the unit of the prenasalised stop, nor does the voicing. In the case of nhx [V|, one can observe the development in the de­ gree of stricture from a (nasal) stop to a fricative. The voiceless prenasalised obstru­ ents are clearly aspirated, again with the exception oí nhx [ŋx]. It is important to note, however, that the oral part including aspiration is not realised by some speakers. Thus nh [n], a voiceless alveolar nasal, is a possible realisation of nht [ŋt], mh [m] of mhp [mp] and ŋh [ŋ] of nhk [ŋk]. In this respect, the properties of voiceless prenasalisation in Mbugu are not different from those in other East-African Bantu languages such as Nyamwezi or Sukuma (see Schadeberg (1992:17) and Maddieson (1991)), and Pokomo (see Huffman and Hinnebusch 1998). Voiceless prenasalisation is common in the area; both Pare and Shambaa share this feature with Mbugu. The prenasalised obstruents occur mainly but not exclusively, word-initially in nouns in class 9/10. Some examples of word-internal prenasalised obstruents are: kimhpanga 'falcon' (I+NMb), kufúnhtu 'to pour' (1Mb), ku'onhti 'to wash' (1Mb), kuhunhta 'to blow' (NMb), Irenhte name of a village (I+NMb), samánhko 'kidney' (1Mb), hinhka 'push' (1Mb), isonhká 'dust' (1Mb), inhkóngolo 'egret bird' (I+NMb), mnhkangwáná name of a clan, ikúnhkuro 'knee' (NMb), jinhka 'jump, fly' (NMb), inhka 'give somebody' (NMb), ihánhka 'basket for food' (I+NMb).

4.3

4.4

Vowels There are five basic vowels: a, e, i, , u. Occasionally, vowels are realised with nasality, as in the word ħũữ 'this' (IMb-Bum). There is some sporadic individual variation between e and i and between  and u. There is no length opposition. A long vowel does occur, as in the roots kaa 'home', makáa 'charcoal' (NMb), inoo 'whetstone' (NMb), nhkoo 'monkey' (I+NMb); and also in the Swahili loans saa 'hour' (I+NMb), taa 'lamp' (NMb), sikúnhkuu 'feast' (I+NMb); and in an optional variant túwáá of the habitual tense marker túwa, for example nitúwáákúru or nitúwakúru T used to cultivate.' Apart from the last example, long vowels can be analysed as disyllabic for various reasons: either because they have falling or rising tones and a disyllabic analysis simplifies the tonal analysis; or because they are the result of two morphemes coming together, as in maasemu 'breasts' (1Mb), va-agirú 'men' (1Mb), va-abá 'fathers' (1Mb); or because in­ flection affects only the second of the two, as is the case with the long vowels in the verbs -sakaa —» -sakáe 'be worn out' (NMb), -ikàa → -íkáe 'sit, reside' (NMb), and others (see section 5.1.5).

104

CHAPTER 4

4.5

Tone There are basically two tones: High and Low. In this work, only High tones are marked, with an acute accent; vowels with no tone mark are Low toned. Contour tones can be analysed as sequences. The tone-bearing units are vowels and the syllabic nasal m that occurs as nominal prefix. The distinctive function of tone is demonstrated by the minimal pairs in Table 4.8. form kái mhpala mkóra iva

tone HL LL HL HL

meaning form tone push káí HH gazelle LH mhpalá in-law mkorá LH steal iva LL Table 4.8: Minimal pairs for

meaning if first tradition doctor forget tones.

language 1Mb I+NMb 1Mb NMb

Tone also plays a role at the non-lexical level, as the pronominal prefixes for noun classes 1 and 3 differ only in tone. There are no restrictions on possible tone sequences. Disyllabic roots have four possible patterns, see Table 4.12 below, while trisyllabic noun roots have eight possi­ ble patterns, see Table 4.9. Thus tone is distinctive on every syllable of the word. However, in verbs restrictions on tone sequences do occur. Lexically, verb roots are either High or Low in tone irrespective of the number of syllables. Such a limitation of possible lexical tonal sequences in verbs is the norm for Bantu languages. tone pattern LLL LLH LHH HHH HHL HLL HLH LHL Table 4.9:

example meaning ki-purukwa nightjar bird i-gorotó back of knee sisírí thin, narrow, short sénéné cricket i-zómólo big spoon i-nhkóngolo egret bird m-hálumá belt i-shamíro anvil Possible tone patterns of trisyllabic

language I+NMb I+NMb NMb I+NMb I+NMb I+NMb 1Mb 1Mb roots

Contour tones occur on long vowels but, as was described above, these are sequences of syllables, for example, fiija [fï:ja] 'finish' (1Mb) which is the causative of/7 'finish'; kirii [kirî:] 'return' (1Mb) and kirí-tí-án-a 'answer' with a causative and reciprocal suf­ fix (see section 5.6). 4.6

Syllable structure Syllables are always open. The only possible coda is the glide y, which may only occur after a, as in chay 'tea', a loan from Swahili, and káy 'three' (1Mb). The

PHONOLOGY

105

rhyme consists of a vowel or a nasal. The only possible syllable with a nasal as rhyme consists of a syllabic m nominal prefix. Long vowels are considered to be disyllabic (see section 4.4). The onset can be empty. If the onset is filled it consists of one con­ sonant or a consonant followed by a glide. The prenasalised stops are considered to be one consonant. Word-medial nasal-stop clusters are syllabified with the cluster in the onset of the second syllable, including the nasal part, e.g., mhpó.mbe 'water' (NMb). If the onset contains a glide, the rhyme does not contain a round vowel. There are no restrictions concerning the first consonant followed by a glide in the onset. The exam­ ples in Table 4.10 show the various types of possible syllables;  = consonant, V = vowel, G = glide, N = nasal. form syllable structure meaning language mhpó.mbe cv.cv water NMb whetstone NMb i.no.o v.cv.v N.CV.CV m.dá.ngu kind of weed I+NMb nhkwe.no CGV.CV chicken 1Mb CGV.CV hwa.'a return 1Mb CV.CGV smell 1Mb hé.fya CGV.CGV fya.fya fastness NMb N.CGV.CGV straw m.fyé.fye 1Mb Table 4.10: Examples of syllable structure The syllable structure is typically Bantu and is valid for both languages, Inner Mbugu and Normal Mbugu. It deviates in at least three important points from Cushitic sylla­ ble structure: 1. the exclusion of closed syllables, 2. the possibility of syllables with no onset, 3. the syllable boundary falls before the nasal-consonant combination. When morphemes combine there are adaptations to achieve this syllable structure. The most prominent ones are vowel coalescence and glide formation. When two vowels come together, it is morphologically determined whether they have to coalesce into one syllable, i.e. one rhyme, or not. First, I restrict myself to those situations where the onset is already filled by a consonant. If the two vowels have to fit into one rhyme, the first vowel is elided, unless it is the high round vowel  which can be pushed into the onset as a glide w as long as the second vowel is not a round vowel. If there is no ne­ cessity for accommodating the two vowels in one syllable, as is the case with nominal prefixes, this glide formation can still take place, but only optionally, and is then also allowed before round vowels. The vowels have to fit into one rhyme when prepositional prefixes (PP) com­ bine with demonstrative and possessive roots and the like, e.g., a-ose 'class 6-all' (NMb) becomes ose. The high vowel  of the prepositional prefixes becomes a glide w

106

CHAPTER 4

before a non-round vowel in nominal pronouns (I+NMb) (see 5.2), e.g.: Iwá (from l­­ cí), vwá (from vú-á), kwá (from kú-á), but not before round vowels, thus ó (from kúó) vóse (from vú-óse) 'class 14.1-all' (NMb), and úlo (from ú-lú-o) 'class 11-that' (NMb). Noun class prefixes containing  also form glides as in mw-ahla fire' (1Mb), lw-atame Tight' (1Mb), but this glide formation is optional and occurs mainly in fast speech, e.g., mu-ezá or mw-ezá long' (NMb), vu-áho or vw-áho 'porridge' (1Mb), and also, occasionally before round vowels as in mw-óshi 'smoke' (NMb).7 The glide formation applies regularly to the high round vowel but not regularly to the high front vowel. Thus the class 5 nominal substitutive forms are lá and not lyá from lí-á and lo (from li-o), and similarly for class l O jo (from ji-o), ja (îrom ji-a). It does, however, become a glide in the substitutives of class 8 with prefix vi resulting in vyo, vya 'class 8'. Note that its singular counter-part class 7 prefix ki has a palatalised form as prepositional prefix, chi. Both i and  become glides if the onset is not yet filled by a consonant and has to be filled because the second vowel takes the rhyme of the same syllable. This can be observed in the substitutives wá (from u-á), class 1 or 3, and yá (from í-á), class 4 or 9, and in nouns such as -asemí 'class 5-female breast'. The glides are only formed when necessitated by restrictions on syllable structure. No glides are formed in the proximal demonstratives úu (from u-u: class 3) and ii (from /-/, class 4) (see 7.4). But the referential demonstratives for the same classes do form a glide as a consequence of an extra vowel that is suffixed to these forms: uwo (from u-[u-o]) for class 3 and iyo (from i-[i-o]) for class 4. Another adaptation to meet the syllable structure when morphemes combine is stop formation and prenasalisation. Pre-nasalised stops are the result of the fusion of a class 9 or 10 nominal prefix with the root-initial consonant. The prefixes for classes 9 and 10 are not syllabic but consist of nasality realised on the noun root-initial conso­ nant. The phonological processes of nasal plus stem-initial consonant integrating into one syllable onset can only be shown from the class pair 11/10 since the pair 9/10 has identical forms in singular and plural (see section 6.3). These processes are the follow­ ing: Stops acquire an extra nasal, voiced before voiced, voiceless before voiceless. Voiced fricatives become prenasalised stops. Nasal initial stems do not undergo changes. The glide  becomes a palatal nasal in class 10, luyungu, pl: nyungu 'cala­ bash plant' (I+NMb). The sound h becomes a voiceless prenasalised bilabial, mhp. The following examples of nouns in the singular/plural class pair 11/10 provide us with evidence for this process. class 11 lupaé ludezu lukoko lugwijo

class 10 mhpaé ndezu nhkoko ngwijo

meaning saliva, curse beard calabash plant post

language 1Mb 1Mb I+NMb NMb

107

PHONOLOGY

class 11 lughúnda luvaju luzoyá luzíghi lurimwá lumuhlú lungóri lunábo luhambá luháwa luhembe

4.7

meaning class 10 ngunda sole of foot mbaju rib ndoyá feather ndíghi rope ndimwá kind of pumpkin plant muhlú rope ngóri steep path nábo rib mhpambá machete mhpáwa ladle mhpembe horn Table 4.11: Nasal + obstruent processes

language NMb NMb I+NMb NMb I+NMb 1Mb 1Mb 1Mb I+NMb 1Mb NMb

Word structure and phonotactics The minimal word is CV. In fact in Inner Mbugu many verbs only consist of

CV. The morphemes that are analysed as clitics, such as sha 'very' (see section 5.1.8), are clitics because they form a phonological unit with the preceding word re­ gardless of its category. However, there is no phonological reason for these to cliticise since other words consisting of just CV contain enough material to be independent. The canonical shape of nouns is prefix-stem, and for all classes except 9/10 the prefix consists of a syllable CV- or V-. The noun root is mono-, di- or tri-syllabic, resulting in words of the shape (C)V-CV, (C)V -CVCV and (C)V-CVCVCV. Monosyllabic nouns in class 9/10 do occur but only in Inner Mbugu, e.g., 'u' 'force', dé 'goat', wá 'cow, cattle'. Noun roots with more than three syllables do occur but they either contain reduplications, e.g., msangalagala 'sand' (NMb), m-dudukéno 'calabash for oil' (1Mb), or they are borrowed from Swahili, e.g., kaskazini 'North' (I+NMb), or they are compounds, e.g., kákárúu 'grandson' and kokorúu 'granddaughter' (1Mb). There exist a number of longer nouns that do not belong to these groups, such as mkalagázo 'Saturday', m-shuzamizi 'woman who has a child' (1Mb), ramburia 'sister' (NMb), nhkulumbiza 'finch' (I+NMb), i-pambarito 'butterfly' (NMb), i-paparisho 'butterfly' (1Mb), i-tusharimi 'caterpillar-like insect' (I+NMb), sangavasi 'ant-like insect' (I+NMb), ki-dóghólosá 'armpit' (1Mb), ki-kongomeru 'Adam's apple' (I+NMb), ki-riandohi 'whirlwind' (I+NMb), ki-kililimo 'mist' (1Mb). Verb forms are usually longer due to a rich system of inflectional and deriva­ tional affixes even if mono-syllabic roots are common. Verbs, like all words, end in a vowel, but the root usually ends in a consonant. The typical shape of a verb root is CVC and an inflected verb typically has the following shape: CV-CV-CVC-VC-V. The vowels that occur in the nominal and prepositional noun class prefixes are i, a, and u, as well as e alternating with a. In affixes in general these three basic vowels are predominant.

108

CHAPTER 4

There are statistical differences in the distribution of consonants in root-initial position as opposed to other positions. In both cases the consonant is intervocalic. Prenasalised consonants are rare in root-initial position8 but common otherwise. The op­ posite can be said about non-pre-nasalised stops, which are common in root-initial position but rare as root-medial consonant, with the exception of  which is not rare in that position. The statistical differences are valid for both Inner and Normal Mbugu. The velar nasal is rare. Intervocalically, it only occurs in Inner Mbugu words: dáŋá 'beastings', aŋila 'above', aŋa 'home', iŋa 'sibling (in compounds)',  'nose', ŋé 'bite', ŋola 'war', and: ŋombe 'cow', ŋuŋunhto 'kind of wind', ŋanya (a place name), ŋambai (a personal name). The distribution of tone is not even. In the following I present the break-up for the tone patterns of mono- and disyllabic noun- and verb roots. In disyllabic noun roots the numbers for the four possible tone patterns are given in Table 4.12, based on a count of all disyllabic noun roots in my dictionary files.9 NMb (including I+NMb) 1Mb total pattern 307 LH 128 179 18 78 HH 60 LL 145 50 195 96 235 139 HL Table 4.12: Tone patterns for disyllabic noun roots. The tonal pattern HH has relatively few attestations. The tonal pattern LH in Inner Mbugu is remarkably frequent and roughly half (85 out of 179) of these end in e. Changing the tonal pattern to a sequence of Lows and a final High tone and at the same time changing the final vowel to e is one of the ways of transposing a word to Inner Mbugu, see 1.7. Some words in my data show tonal variation: LL or LH (twice), LL or HL (thrice), LL or HH (twice), LH or HH (once), and HL or HH (twice). I have been un­ able to check these and so these may be mistakes in the data. Mono-syllabic noun roots need a high tone in Inner Mbugu, but not in Normal Mbugu. The only exception is the Inner Mbugu word kwa 'path' which is a short form of kwahli. The numbers for the tonal patterns of mono-syllabic noun roots are pre­ sented in Table 4.13. pattern

1Mb NMb (including I+NMb) 1 13 H 30 26 Table 4.13: Tone patterns for monosyllabic noun roots

"L

The verb roots in Normal Mbugu are all either High or Low. The verb root consists either of a consonant and a tone (realised on the following vowel) or of CVC. The

PHONOLOGY

109

tone on derivational suffixes is predictable from the tone of the root and that of the inflection. The distribution of High versus Low verb roots is more or less balanced. Table 4.14 contains the number for each tonal type for verbs in Normal Mbugu, in­ cluding those that are common to both languages. pattern  CVC H 7 74 L 5 116 Table 4.14: Tone patterns for Normal Mbugu verb roots The Inner Mbugu pattern is not at all comparable, see Table 4.15. In Inner Mbugu there is a high number of verb roots that are not of the type C- or VC- but that end in a vowel that is part of the root (and not inflectional) and this fact makes the compari­ son difficult. As a consequence of the fact that Inner Mbugu verbs may end in a noninflectional root vowel there is room for more tonal distinctions and this room is used. In Table 4.15 we have over-simplified matters by dividing the verbs into those ending in (inflectional) a and those ending in a vowel (not a) that is part of the root (but see 5.1.5). pattern final vowel is not a final vowel is a 41 20 "H 0 L 0 HL 67 34 LL 12 14 HH 4 2 LH 22 4 1 able 4.15: lone patterns for Inner Mbugu verb roots The absence of Low tone in monosyllabic verb roots is noteworthy. The same restric­ tion is valid for noun roots. Thus, in Inner Mbugu, monosyllabic lexical roots of the major categories have a High tone, a restriction that is not valid for Normal Mbugu. The HL and LL tone patterns are as expected since the (inflectional) final vowel should be Low in a standard Bantu verbal system. The few verbs with HH tones can be dismissed as exceptions and similarly the four LH verbs ending in a. The twentytwo LH verbs ending in a non-inflectional vowel are a-typical for Bantu since the final (inflectional) vowel should be low and in Bantu there is a tendency to have only H or L tone verb classes. Thus, the restriction in Normal Mbugu that verb roots only have one tone does not hold for Inner Mbugu. As far as tone rules are concerned, the inflectional prefix that precedes subject prefixes in the optative and negative tenses consists of just a H tone. The realisation of this high tone is on the following syllable, in this case the subject prefix. In contrast, the floating High tone preceding the extra final ye of the perfect is realised on the pre­ ceding syllable, i.e. the penultimate. The High tone of a subject prefix in the verb may

110

CHAPTER 4

spread to the next syllable depending on which tense prefix follows; some tense pre­ fixes seem to be optionally susceptible to this tonal spread (see 5.1.3). The tone of the stem spreads to the penultimate syllable. A similar tone spreading rule was observed on the optative final vowel e: the high tone of the verb stem spreads to this final vowel e (if present) as in àdósé 3SG:follow (NMb) but in this tense the verb stems with a non-inflectional final vowel also trigger tone spread. These morphophonologival tone rules are discussed again in the chapter on verbal inflection.

5

THE VERB

The main body of this chapter consists of a discussion of the numerous "tenses" of the verb, the affirmative "tenses" (sections 5.2, 5.3 ), and the negative "tenses" (section 5.4).1 First the essentials of the structure of the verb and the prefixes other than "tense"-prefixes are discussed briefly in section 5.1. Following the discussion of the "tenses" of the verb, I look at the copula and verbs 'to be' (section 5.5), and verbal derivation (section 5.6). The differences between the verbal morphology of Inner Mbugu and Normal Mbugu are minimal. All differences are highlighted but the verb morphology of the two languages is presented simultaneously. 5.7

Overview The verb consists of a stem and a series of inflectional prefixes. The stem may also contain derivational suffixes (see section 5.6). The final vowel of the verb is ar­ guably not part of the stem; rather it is an inflectional vowel (see section 5.1.5). With the exception of imperatives (see section 5.1.7) and infinitives, which are in fact class 15 nouns, every verb form contains a subject prefix, which is usually the first mor­ pheme of the verb. A verb form may also contain one or two object prefixes. After the subject prefix there can be up to three prefixes within a wide semantic domain encom­ passing inter alia tense, aspect, mood, and polarity; it is these that I call the "tense" prefixes. The position before the subject prefix can be taken by a prefix expressing negation and by a high tone marking the boundary of a predicated domain. The inflec­ tional paradigms or "tenses" of the verb are certain combinations of these prefixes (all but subject and object) and the final vowel. There are many inflectional paradigms, and the list presented in Table 5.2 is probably not complete since every new text col­ lected tends to result in one or more new inflectional paradigms. Richness in "tenses" of the verb is not uncommon among Bantu languages in the area, see, for example, the extensive list in Roehl (1911: 108-151) on Shambaa. The structure of the verb can be summarised as in (5.1): (5.1) (neg)-(H)-subject-"(tense(tense(tense)))"-(object(object))-stem-infl. vowel The morphemes that can fill the subject slot and the object slots are given further down in Table 5.1; those that can fill the "tense" slots are given in Table 5.2; those that can fill the inflectional (final) vowel (infl. vowel) slot are given in Table 5.4. The negation slot can be filled by the negative prefix te-. A floating high tone, H, before the subject prefix, must be posited for certain "tenses". Some example sentences2 taken from texts are given below to illustrate the structure of the verb. Examples (5.2) and (5.3) show the use of an object prefix há and li respectively and example (5.5) illustrates that a verb form can contain two object prefixes, in this case chi 'it' and m 'him/her'; examples (5.4) and (5.6) show an inflec-

112

CHAPTER 5

tional vowel that is different from the default low-toned a: example (5.4) shows the use of a negative prefix and the high tone, which is marked on the subject prefix, indicating the predicated (negated) domain; example (5.7) shows once more the validity of the stem minus inflectional final vowel as the basic lexical unit within the verb forms since it is precisely the stem that is replaced by the dummy filler nooni, which is used when the speaker cannot find the right word. This filler in (5.7) is preceded by a subject and "tense" prefix and followed by the inflectional final vowel a. (5.2) n-ée-ka-há-so ana (1Mb) 1SG-PST-CSEC-16-go 16.home 'I had gone home.' (5.3) tu-ku-za-li-ingia (NMb) lPL-COND-lT-5-enter 'When we entered it (the land =class 5),' (5.4) te-vé-na-ghuhwá (NMb) NEG-2-EXPEC-marry:NEG 'They are not married yet.' (5.5) u-ka-ta-chi-m-lot-ija 2SG-CSEC-EVI-7-1 -point-CS 'If you point him thus,'

uvu (NMb) 14.1 :DEM 1

(5.6) mhaka w-ó-manj-e (NMb) border 2SG-REL-know-PF. 'The border that you know' (5.7) i-ku-nooní-a ... í-ku-gér-a (IMb) 9-COND-"dummy"-F 9-COND-bring-F 'If it -whatmething ... if it brings' 5.1.1 Vowel coalescence, i-initial stems, and morphotonology Vowel coalescence rules (see section 4.6) apply in two domains of the verb: they apply in the domain preceding the object prefix slot and in the domain following the object prefix slot but the boundary preceding the object prefix behaves phonologically like a strong boundary. Thus, in the first domain, vowel coallescence applies be­ tween subject and "tense" prefix in (8), (9) and (11), and between the object prefix and the stem and between the stem and the final vowel in (10). But it does not apply across the strong boundary preceding the stem in (12), (13), (14), (15) and (16). The relevant sequences are underlined; the strong boundary is marked by +. (5.8) ni-áa+kwáha

→

náakwáha

'I am tired'(1Mb)

THE VERB

(5.9) tu-áa+kwáha

→

twáakwáha

'we are tired'(1Mb)

113

(5.10)é+há-áho-íe

→

éháhóye

'he saw it (the place, cl. 16)' (1Mb)

(5.11) é-áa+í-kícha

→

áíkícha

'he found them (the trees, cl. 4)' (NMb)

(5.12)u-si±iva-é

→

(5.13)vé±áhóíka (5.14) i±áho (5.15)'-tu±á-bó'i

→ → →

usiivaé

'don't steal'(NMb)

véáhóíka

'they are visible'(1Mb)

iáho

'it (the leopard, cl.9) sees' (1Mb)

túábó'i

'we should make them (cl.6)' (1Mb)

(5 A6) ni-té±á-vonshwa → nitéávónshwa 'to me they (cl.6) were shown' (NMb) A number of Normal Mbugu verbs have a stem-initial i which is lexically determined and has no discernable semantic content. From a lexical point of view it can therefore be analysed as part of the root. Such an analysis does, however, entail phonological and morphological problems, as this would run counter to the general verb root struc­ ture which is  or VC.3 These verbs are ika-a 'sit, stay', itang-a 'call' (more often used in its passive itang-w-a 'be called'), idim-a 'can, be able', inhk-a 'give' (and its causative inhk-ij-a 'take out'), and izu-a 'fill'. This stem-initial i is sometimes re­ placed by an object prefix, e.g., túmnhke 'let us give him', and the i may even precede the object prefix, e.g., tuimnhkwa 'we are given him'. The i may also undergo elision when preceded by a subject prefix, e.g., edima 'he can', éenhkiwe 'he was given'. The i is part of the lexical stem, however, and not an object prefix since it occurs when no object is present and when no object can be added, e.g., váaIkáa naye 'they stayed with him'. The same lexical items show the same stem-initial i in related Bantu lan­ guages such as Pare: iŋka, hanget, íkaa, ídima, ízua, and Shambaa: inka, itanga, ikala. No Inner Mbugu verb has such a stem-initial i. Other verb-initial stems, also in z, do occur, e.g., iv-a 'forget; steal',4 both in Inner and Normal Mbugu, and the i in this case forms an integral part of the stem. The floating high tone before the subject prefix indicating the predicated do­ main is realised on that subject prefix in certain "tenses", as in (15). The floating high tone preceding the extra final ye of the perfect is realised on its preceding syllable, i.e. the penultimate ((10)). A number of "tense" prefixes can receive the spread of a high tone of the preceding prefix, usually the subject prefix ((17)). Some "tense" prefixes seem to be optionally susceptible to this tonal spread ((18)). (5.17) é-ze-kúru → ézekúru

'he went and cultivated' (1Mb)

(5.18) mw-á-ra-áho (→ ) mwárááho

'do you (pl) see (it)?' (1Mb)

114

CHAPTER 5

5.1.2 Subject and object prefixes The noun class subject prefixes as well as the object prefixes differ from the pronominal prefixes only in the high tone that they carry. Pronominal prefixes are used in demonstratives, possessives and on other nominal modifiers that agree with the head noun. The prefixes of class 1 (human beings, singular) are not identical to the pronominal prefix, that is, they have a or e for the subject prefix and a low toned m for the object prefix which becomes mu (mw) before vowels. The noun class subject pre­ fixes of classes 1, 2, 6, 12 and 16, i.e. those classes that contain an a in the pronominal prefixes, have an e instead of an a in most "tenses"; the object prefixes do not show this alternation (see section 5.1.6). The prefixes for first and second person carry low tone with the exception of the second person plural object prefix. The subject and ob­ ject prefixes are identical to those in Pare, including the remarkable a/e alternation. The only difference is that class 7 is ki in Pare and chí in Mbugu (Kotz 1909: 51-2; Kähler-Meyer 1964: 100). In my data I sometimes have ú instead of lú as object prefix for class 11 and this is probably interference from Swahili which has u. A recent bor­ rowing from Swahili and Shambaa is the introduction of an extra noun class for ab­ stract nouns in u- which I indicate as 14.2 since it has the same historical origin as class 14.1 vu; class 14.2 has not been attested in Pare. The subject and object prefixes are presented in Table 5.1. The non-speech act participants are divided into the various noun classes (see section 6.1); 1 indicates third person class 1, 2 indicates third person class 2, etc. Classes 1 and 2 contain human third persons in singular and plural respec­ tively. I have added the reflexive marker to this table because it occurs in the position of the object prefix. person / class l SG 2SG 1PL 2PL

1 2 3 4 5 6 7 8 9 10 11 12 14.1

subject prefix ni  tu m (u) é/á vé/vá ú í li é/á chí vi i

ß lú ké/ká vú

object prefix ni ku tu m(ú) m (u) vá  í li á chí vi i

ßr

r

lú/ú kå vú

115

THE VERB

person / class 14.2 15 16

subject prefix object prefix ú ú kú kú hé/há hå REFL kú Table 5.1: Subject and object prefixes

The object prefix is used when the object head noun is understood ((11), (15), (16)) or when it is definite ((19)). Double object prefixes are possible'but not common (see (5) and (20)). (5 9) é-háhóye hódi to ni ha-kuhlø (1Mb) l-16:see:PF place there COP 16-good 'He saw the place was good.' (5.20) vé-ka-há-ní-'ó (1Mb) 2-CSEC-16-lSG-give 'They gave me the place.' The reflexive object pronoun is kú for all objects-cum-subjects. Examples (21) and (22) show the use of the reflexive object pronoun; the (a) examples are Inner Mbugu, the (b) examples Normal Mbugu. (5.21) a. ú-kú-hapáre gomaé 2SG-REFL-close:SBJ clothes 'You should wear a blanket.' (5.22) a. é-tá-kú-hlómi mahlóma 1 -EVI-REFL-anoint oil

b.

b.

ú-kú-ghubike shuká 2SG-REFL-close:SBJ clothes é-ta-kú-víra mafuta 1 -EVI-REFL-anoint oil

'She anoints herself.' 5.1.3 Other verbal prefixes, "tense" and polarity The prefixes that can precede the subject prefix are the negation marker te (si for the combined negation and first person singular subject) and a floating High tone which is realised on the next vowel, i.e. the subject prefix, and which is used for all the non-assertive "tenses": the subjunctive, conditional, and negative "tenses".5 Table 5.2 summarises the "tense" prefixes that occur between the subject prefix and the object prefix or the stem. A more precise statement on the meaning of these prefixes is only possible after examining the various "tenses" in which they combine with each other and with the final vowel. In addition to the negation marker te men­ tioned above there are are also several negative prefixes that fit in the slot of "tense" prefixes.

116

CHAPTER 5

form å åa/ée ta te na ne re/ra za/ze ka ka kú hé tuwáa chéri réngé manga si se såa sée

gloss INC PST EVI NARR C.EXPEC EXPEC BGND IT CSEC NEG.FUT COND PST.COND HAB STILL HORT SPEED NEG NEG NEG NEG

name incidental past logical sequential (gloss is EVI for evidential) narrative counter-expectational expectational background itive (thither) consecutive negative future conditional past conditional habitual unfinished action hortative immediate action negation negation negation negation Table 5.2: " Tense " prefixes

Most of these prefixes also occur in Pare or Shambaa. They contribute meaning in the Tense-Aspect-Mood domain, with the prefixes se, si, såa and sée also indicating po­ larity, and the semantic domain in which these prefixes operate is even larger, cf. ex­ pectational, immediate action, etc. The polysyllabic prefixes show some variation: tuwáa or túwa; réngé, rångå, ronga, or ronge. These polysyllabic prefixes are recent incorporations into the inflectional verbal system. The itives za/ze6 and ha indicate movement, 'go and ...', and the attitude markers ne, na, ta, te, relra are usually added to other "tense" prefixes, although they can all occur as the only "tense" prefix in a given verb form. These itives and attitude markers occasionally occur after the object prefix. This was never found in texts but elicitation provided evidence that they can take that position (see example (23)). This positional capacity, if truly valid, is the only formal characteristic that sets these prefixes apart from the other "tense" prefixes. (5.23) ni-vá-ku-ta-áyja (IMb) lSG-2-2SG-EVI-show 'I should show them to you.'

THE VERB

117

5.1.4 Relative marker The relative marker  has been attested, albeit marginally, in the position pre­ ceding the stem as in (24). The common position of the relative marker, however, is in the relative pronoun preceding and often cliticised to the verb (e.g. (25) and (26)). I have only two examples in my data where the relative marker ó is verb internal and both involve the irregular verb 'to be' (as in (24)). Pare uses the constructions as in (25) and (26) with relative pronoun introducing the relative clause, and also in (24), which would be handu ehó vého in Pare. Swahili has an internal relative marker  as one of the possible strategies. It is possible that example (24) is modelled after Swa­ hili. (5.24) hanhtu ve-o-áho (NMb). place 2-REL-be 'place where they were' (5.25)j-ó-j-ée-z-íe (NMb) 10-REL-10-PST-arrive-PF 'that has arrived' (5.26) l-ó-n-'óró-ye (1Mb) 5-REL-lSG-close-PF 'that I closed'. Verb forms in relative clauses are not different from other verb forms. The use of a relative pronoun is not obligatory. It is remarkable, however, that in relative clauses the subject prefix often contains an a instead of the usual e (as in (27)) and often the verb form has a PF-suffix. A very common construction is one of a (relative) clause preceded by ni-h-ó COP-16-REL 'It is that ...'. All "tenses" that occur in independent clauses also occur in relative clauses, such as perfect, expectational, consecutive, con­ ditional, past imperfective, past perfect, past, "not-yet", and the default-"tense". (5.27)hé-na mnhú á-tikwa ... a-ka-tikwa (NMb) 16-with being 1-lifted 1-CSEC-lifted 'there is somebody who was lifted and he was lifted' 5.7.5 The inflectional final vowel All verbs end in a vowel. For most "tenses" and in the infinitive this vowel is a; for some "tenses", notably the subjunctive and the perfect and some negative "tenses", it changes to e or iye. Thus, the final vowel can be analysed as an inflectional vowel and not as belonging to the stem, being predicted by inflection and having a as the morphologically unmarked value. Derivational suffixes generally have a -VC shape and are added to the stem. The inflectional final vowel follows the derivational suf­ fixes. This is the common practice in Bantu languages. There are, however, exceptions

CHAPTER 5

118

to this pattern of verb sterns ending in a consonant that is obligatorily followed byan inflectional final vowel. Some verbs end in a vowel that never changes, and that is thus not inflectional but part of the stern. Oerivational forms of such verbs, however, end in an inflectional vowel. This behaviour of the exceptional cases is not different from other Bantu languages. In code-switching involving Bantu languages as the matrix language, embedded verbs never show verb-final inflection except when a derivational suffix is added, see Myers-Scotton (l993a:80 (ex. 1),87 (ex. 10), 109; 2000), Oe Rooij (1996: I 06-7); Swahili has many verbs from Arabic that end in a noninflectional u or i, but not a;7 their derivational forms, however, end in the inflectional a again. However, there is a difference between Inner Mbugu (not Normal Mbugu) and this "standard" Bantu situation: While in Swahili, for example, such exceptional non-inflecting final root vowels are never a, Inner Mbugu (not Normal Mbugu) does contain verbs ending in an a that"never inflect, compare the first and the second series of verbs in Table 5.3 below. 8 The proportionally high number of verbs ending in a non-inflecting vowel is also remarkable. language infinitive meaning I SG:SUB kukti'ti take away nfkti'a 1Mb open 1Mb kuhla nihlit kubvti build nibva 1Mb kltwahit drink niwaha 1Mb kuayjti show nfayja or niäyje 1Mb kujikit bury nfjikit or nfjike I+NMb kuputa smoke nfpute 1Mb kujuta pull nijute I+NMb kuv6kti start nfvoke I+NMb Table 5.3: Verbs with non-injlectional and injlectional final vowel a in 1Mb The inflectional verb endings are presented in Table 5.4. The high toned i and the negative final floating high tone, only affect the verbs that have an inflectional final vowel. If a verb ends in a double aa in the citation form, the last a changes to e and the high tone is realised on the penultimate a: usiivae 'Oon't forget!' (NMb), tekaghatae (NMb) 'He won't get tired'. function (default) H-(y)e PF perfect PF perfect SB] subjunctive NEG some negative "tenses" NEG some other negative "tenses" Table 5.4: Injlectional final vowels

form

gloss

a

F

i+ tia e

e

H

THE VERB

119

The additional final element ye and the high tone on the penultimate syllable apply to all verbs. The perfect suffix is not iye but e after a derivational suffix such as passive w, applied i, ir, causative ish, ij, stative ík. The perfect is also e after a non-inflectional vowel e as is evidenced by the perfect of the verb he 'to put ' in (28). The perfect suf­ fix is -ye after non-inflectional vowels such as o, i, u, and after non-inflectional a. The perfect suffix is -e plus stop-formation after a palatal nasal as in chifunyanįe in (29) and similarly with the verb ku-manya 'to know' (NMb). Exceptional behaviour is found with another verb with a final nasal, vona 'to see' (NMb): the perfect suffix has a realisation following and preceding the stem final consonant yielding vwene (as in (30)). Such infixation, or imbrication, is known from other Bantu languages (Bastin 1983) but in Mbugu it is limited to this case. (5.28) chóe-ké-é ata (IMb) 7:REL 1-put-PFhere 'What he has put here' (5.29)chifunyanįe (NMb) chi-funyany-e 7-come out-PF 'It came out, it happened' (5.30)twéevwene (NMb) tu-ée-von-e lPL-PST-see-PF

'we have seen' After verb stems that have a long vowel, sometimes both vowels change (as in (31a)) but often only the last vowel is replaced (as in (31b)) and in vekweye 'they climbed' (NMb) from the verb kwea. Compare also the perfect in the in ninythe 'I drank' from the verb nyá 'drink' and in uzihe 'it (= month) to come' from the verb za (NMb) 'come' but iziye 'it (=week) to come' from the verb zó (IMb) 'come'. (5.31) a.

véeikée(lMb) vé-ée-ika-íe 2-PST-stay-PF 'they stayed'

b.

vasituikaíye (NMb) va-si-tu-ika-íe 2-NEG-1 PL-stay-PF 'they didn't stay with us'

The element ye is not the only post-final element. Verb forms may be followed by question clitics (see section 5.1.8). 5.1.6 The a/e alternation This vowel alternation does not occur in areas outside the system of verbal pre­ fixes, and among the verbal prefixes, it only affects subject and "tense" prefixes and

120

CHAPTER 5

not object prefixes. This alternation is thus morphologically restricted and is not de­ termined by the phonological environment. As regards the subject prefixes, the allomorphy is neatly conditioned by the presence or absence of certain "tense" markers and the condition is uniform, that is, the same condition holds for the same alternation in all the relevant subject prefixes; these are the subject prefixes of classes 1, 2, 6, 12, 16. Those classes that contain an a in the object prefix, or in any other concord prefix such as the pronominal prefix, have an e instead for the subject prefix of most "tenses". In addition, the class 1 subject prefix, i.e. third person singular human, is e for these same "tenses". These subject prefixes have a in the subjunctive and before the conditional he. These classes are precisely those that contain an a in their subject concord in most other Bantu languages. Two points must be stressed in order to put these statements in perspective. First, before the vowel-initial "tense" prefixes áa, ée and á, the vocalic quality of the subject prefix cannot be determined since this vowel elides due to the coalescence rule. Second, there seems to be some variation possible in actual usage. In texts, the subject prefixes containing e are occasionally heard with a instead, and some contexts, such as relative clauses and the conditional ku, seem to favour a over e; on the other hand, those "tenses" that require a subject prefix contain­ ing a never have an e. A number of these occurrences of a where e is expected can be attributed to interference from Swahili; Swahili has a in these prefixes. Tense prefixes show the same a/e vowel variation as the subject prefixes de­ scribed above, but here the distribution is only partly complementary. With the excep­ tion of the narrative and the negative future ka, all "tense" prefixes that contain an a have a counterpart that contains an e. Each of the pairs áalée, zalze, ralre, talte, naine, sáalsée, kusalkuse behaves differently and hardly any uniform statement regarding the distributional characteristics can be made, or it would be that in non-verbal sentences the a-allomorphs are excluded. The variation is not balanced. The prefixes ta and za are more common than te and ze, while re is more common than ra. In negative verb forms ne is excluded and only na occurs, with a specific meaning of '(not) yet'- Nega­ tive verb forms have  to the exclusion of åa, re to the exclusion of ra, and ze to the exclusion of za (which is the opposite of the distribution in positive verb forms). This is, however, not a general tendency towards e in "tense" prefixes in negative verb forms since te is less common than ta in these forms. In fact, ne and na as well as te and ta are analysed as different morphemes on the basis of their semantic differences, whereas zalze, ra/re and áalée are considered as pairs of allomorphs. The details of this will be clarified in the discussion of the respective "tense" morphemes, see also Mous (1997) on this issue. The vowel alternations of the disyllabic "tense" prefixes show a tendency to paradigmatic vowel adjustments as the "tense" marker réngé HORT has ranga among its variants.9 The negative prefix te which precedes the subject prefix does not alternate.10 The final vowel of the verb form is inflectional and the possible (inflectional) vowel quality values are a or e. The alternation in the prefixes is possibly historically related to this inflectional final vowel e (see Nurse 1979:489-495). This a/e alternation in verbal prefixes is a

THE VERB

121

phenomenon that is limited to Pare, Shambaa, Davidą as it was spoken in the Shambaa area (Woodward 1913/14:92), and Mbugu, and does not occur in other East African Bantu languages. 5.1.7 Imperatives Imperatives consist of the bare verb stem, e.g., kúru mbuvá 'Cultivate the field!' (1Mb). Plurality of the addressee is indicated by a suffix -ni and a high tone on the preceding syllable, e.g., kúrúni mbuvá. The object follows the imperative verb. The imperative of// 'come' (1Mb) has a deviant imperative form, lila. 5.1.8 Clitics A verb form can be made questioning 'what?' or 'how?' by adding the clitic -ze (NMb), e.g., itaghendaze 'what should it do?', cf. itaghenda 'it should do' (NMb). The IMb counterpart is the clitic -mo, e.g., véekúsámo 'what did they do?' (1Mb), véeku-kúrumo 'how were they cultivating?' (1Mb).11 Another questioning clitic is -lé 'where', e.g., usóle 'where are you going?' (1Mb), umhehláyelé 'where did you meet him?' (1Mb). Other such clitics are the intensive extension (see section 5.6), and the plural addressee suffix -ni which is treated in the preceding section. 5.2

The "tenses" In this paragraph all attested "tenses" will be presented. The emphasis is on form; the semantic analysis is rudimentary. All deviations in form from the default will be noted; I consider the default inflectional final vowel to be a, and for those sub­ ject prefixes that show a/e alternation I take e to be the default form. The choice of the default form, a for the inflectional final vowel and e for the subject prefixes, is deter­ mined by the possibility to formulate conditions for use of the other allomorphs most efficiently. The positive tenses are summarised in Table 5.5 and presented in a list be­ low. lable of "tense" default tense past tense resultative completive past past imperfective incidental optative expectational logic sequential narrative background itive

structure SC-0-stem-a SC-aa-stem-a

SC-O-stem-i-ye SC-ée-stem-í-ye SC-ékú-stem-a

SC-á-stem-a -SC-O-stem-e

SC-ne-stem-a SC-ta-stem-a SC-te-stem-a SC-re/ra-stem-a SC-ze-stem-a

122

CHAPTER 5

lable of "tense" structure consecutive SC-ka-stem-a '-SC-kú-stem-a conditional imperfective conditional -SC-he-stem-a habitual SC-túwa/tifwáá-stem-a unfinished SC-chéri-stem-a hortative SC-renge-stem-a

Table 5.5: The positive tenses The verb forms in the list are presented by using the first and third person singular (class 1) of the verbs ku-kúru 'to cultivate' (1Mb) and ku-do-a 'to touch' (1Mb); and ku-im-a 'to cultivate' (NMb) and ku-dós-á 'to follow' (NMb). These forms show for both languages the possible values for the different variables: inflectional final vowel or not (kuru vs. the others), low and high tone subject prefix (lSG vs. 3SG), low tone and high tone verb stem (kudoa (1Mb) and kuima (NMb) vs. kukúru (1Mb) and kudósa (NMb) (see Table 5.6). low tone subject/high tone verb (1Mb) low tone subject/high tone verb (NMb) high tone subject/high tone verb (1Mb) high tone subject/high tone verb (NMb) low tone subject/low tone verb (1Mb) low tone subject/low tone verb (NMb) high tone subject/lowtoneverb(1Mb)high tone subject/low tone verb (NMb) Table 5.6: Structure of the examples for each "tense " in the list The relevant verb forms are presented in a list following this table. When the data for any of these combinations are not available, I have used other subjects or other verbs that are of the same tonal type and in those cases the meaning is given in the notes fol­ lowing the table. Note that in many cases, especially for NMb, I have gaps in my documentation of the forms. Given the large amount of "tenses" and the preliminary state of knowledge about their usage, the labels for the "tenses" should be taken as tentative. Notes to the 1st and a brief discussion of the "tenses" follow after the list. The forms of the positive 1Mb default tense 0...a n ikuru ékúru n ido a édoa past tense áa...a náakúru áakúru

tenses following the scheme above: NMb édósá éima áafwá1

THE VERB náadoa náagwa2 áadoa áaima resultative Ø...-í-ye nikúrúye nikúndíye3 ékúrúye ékúnidíye nidoiye nilukiye4 édoiye completive past ée...-í-ye néekúrúye éekúrúye néedoíye néeangíye5 éedoíye áaimáa past imperfective ékú...a nékúkúru nékúíma incidental á...a nákúni áhúnduka6 áima optative

'-Ø...e níkúru ákúru nídoe ádoe expectational ne...a nnekúru énekúru nnedoa énedoa logic sequential ta...a nitakúrn étakúru nitadoa étadoa narrative te... a nitekúru étékúru nitedoa étédoa

ádósé nííme áime tunehúnduka7 nineima véneima

étaima tutedmdika niteima vétéímá

123

124 background re/ra...a nire/akúru ére/akúru nire/adoa éreladoa itive ze...-a nizekúru ézékúru nizedoa ézédoa consecutive ka...a nikakúru ékakúru nikadoa ékadoa conditional kú...a níkúkúru ékúkúru níkúdoa ékúdoa imperfective conditional níhekúru áhekúru nihedoa áhedoa habitual túwa/túwáá...a nituwáákúrn étuwáákúru

CHAPTER 5

nire/adósa vére/adósa

ázepáta8 ázeima nikalá9 ékalá

tukuvóna vékuvóna he...a

étúwaima unfinished chéri...a nicherikúr échérikúru nicheridoa échéridoa hortative renge...a niréngékúru érengékúru

échériima

árongeima Examples with other verbs used in the list: 1SG:die, 2lSG:fall, 3lSG:want, lSG:plait:PF, 5lSG:walk 63SG:leave, 7lPL:EXPEC:leave, 83SG:get, 9lSG:eat, 10lPL:see.

4

THE VERB

125

Default-" tense": A verb form with only a stem including the unmarked final vowel a and a subject prefix is well-formed and finite. This default "tense" is used for general truth statements, for habituais, and in stories when the time reference has already been established. The same "tense" exists in Pare. I give some examples of the default tense from a NMb text on the history of the Mbugu ((32)-(34)). This default-"tense" construction is particularly used after niho COP-16-REL 'it is that' ((33)-(34)). (5.32) ha-Iá hé-ti-wa hé-lu-ngorí lw-a nhkandu (NMb) 16-DEM3 16-call-PAS 16-11-slope 11-of Nhkandu 'That what is called the Nhkandu Hill.' (5.33) ni-h-o tu-baja (NMb) COP-16-REL 1 PL-split 'We split up.' (5.34) ni-h-ó va-kirindi vé-za (NMb) COP-16-REL 2-Kilindi 2-arrive 'The Kilindi arrived/ Past "tenses": The past ''tenses" make use of the perfect suffix, of the "tense" prefix åa or é or of a combination of the two with "harmony" between them. In addition, there is a past imperfective marker ékú, which is a fusion oféand the conditional kú, which is treated separately below. The "harmony" between åa/ée and the perfect suf­ fix entails that the prefix ée combines with the -i-ye perfect suffix and the åa prefix combines with the åa perfect suffix.12 This åa perfect suffix can only occur in combi­ nation with the åa prefix. The general perfect suffix is i-ye. The past tenses are sum­ marised in Table 5.7. tense prefix åa ée åa ée éku

tense name inflectional suffix a past a past åa completive í-ye completive í-ye resultative a imperfective past Table 5.7: Past tenses

Past áa/ée: The past tense prefix is åa with a falling tone. The vowel of the subject prefix coalesces with the "tense" marker åa- and as a consequence the tone differences between subject prefixes are levelled out, as is the alternation a/e in subject prefixes.

126

CHAPTER 5

The difference between Past and Incidental "tense" is tonal plus concomitant vowel length: for the former a falling tone and for the latter a high tone, cf. náaima 'I cultivated' vs. náima ' I cultivate', áaima 's/he cultivated' vs. áima 's/he cultivates'; the same applies to the difference between Past "tense" and Zero-"tense" forms for class 1/2 subjects (the most frequent subjects): éeima 's/he cultivated' vs éima 's/he cultivates'. The "tense" prefixes áa and è are variants; áa is much more common than è. Past tenses of verbs 'to be' with a nominal complement always have è and never áa. The equivalence of áa and è is exemplified by the following two very similar sen­ tences ((35a, b)) which were uttered immediately after each other in elicitation. It also follows from the fact that both áa and è are used as the equivalent of Swahili me in elicitation ((36),(37)). (5.35) a.

ée-hátisiya (I+NMb) 1 :PST-sneeze 'He sneezed'

b.

áa-bigha miyayú (NMb) 1 :PST-hit yawn 'He yawned'

(5.36) Sw:

wamevuta kamba v-ée-húru lummhlú (1Mb) 2-PST-pull rope 'They pulled the rope.'

(5.37) Sw:

ameuza kibuyu chake áa-dí kishaghú kú'u (1Mb) 1-PST-sell calabash his 'He sold his calabash.'

The past "tense" marker áa/ée is used for states and actions in the past ((38)-(39)). The past "tense" may indicate a resultant state for inchoative verbs (as in (40)-(42)). (5.38) tw-ée-lo na mahloma (1Mb) 1 PL-PST-have with fat 'We had fat.' (5.39)n-ée-te-ómba mzúngu mu-wééséwé maja sandres (IMb) ISG-PST-EVl-request European l-call-PAS:PF major S 'I have requested a European-called Major Sanders.' (5.40) koré y-áa-hú (IMb) pot 9-PST-full 'The pot is full.'

THE VERB

127

(5.41) tw-áa-kwáha (1Mb) lPL-PST-tired 'We are tired.' (5.42) w-áa-hléiwa (1Mb) 2SG-PST-late:PAS 'You are late.' Pare has a verb "tense" a- but there it indicates a commencing or ended act. Shambaa distinguishes short and long a in the "tense" system: a for the dependent distant past, and aa for the independent distant past, see Besha (1989:208). Resultative -í-: The resultative "tense" marking consists of the inflectional suffix -iye. The same suffix is used in the formation of the two "tenses": the resultative and the completive. The high tone of the suffix is realised on the vowel preceding ye regard­ less of whether this is a non-alternating final stem vowel or the i of the suffix that re­ places the inflectional a. Thus nifiíjíye 'I have finished', from kufiija (IMb) with an inflectional final vowel, but also nihliye 'I have refused', from kuhli (1Mb) with a noninflectional final vowel; and also nihláye 'I have opened', from kuhlá (1Mb) with a non-inflectional final vowel a. The resultative is the most common "tense" when expressing states and actions in the past. The term resultative has been chosen because it is used to indicate the re­ sultant state of action verbs (as in (43), (44)). The same resultative meaning has been observed about the past "tense" åa/ée (cf (41), (42) above). (5.43)mweta ú-zúw-e mhpombe (NMb) 3:river 3-fill:PAS-PF 9: water 'The river was full with water.' (5.44) ni-sé-we Wiliam (1Mb) lSG-call-PAS:PF William 'I am called William.' Fixed expressions such as uzihe or iziye 'the one that comes' for 'coming month' and 'coming week', respectively, consist of a resultative "tense" form of the verb za (NMb) or zó (1Mb) 'come'. Both Pare and Shambaa have a corresponding "tense" with the final -iye. In both languages there is a more elaborate rule which establishes the actual from of this inflectional final element, depending on the number of syllables of the stem and on the final stem consonant. Completive Past åa...áa / ée...i-: The completive past is a combination of a past "tense" prefix åa or ée and a suffix -åa or -í-, where åa combines with -åa and ée

128

CHAPTER 5

with -í-ye. Only passive verb stems allow a suffix -(iy)e in combination with "tense" prefix áa, (45). The final -áa which can only occur with a "tense" prefix áa lays em­ phasis on the completion of the action, e.g., yáazakitáa 9:PST:finish:PST 'It is com­ pletely finished.' (1Mb). This "tense" can also express a simple past meaning, (46). (5.45) ma'í y-áa-wáhá-we ni 6:water 6-PST-drink-PAS:PF by 'The water was drunk by cows'

wá (1Mb) 10'cattle

(5.46) badáaye ku-pata mateso v-ée-dumú-ye ní-há-ka i'i(1Mb) after 15-get 6:problems 2-PST-want-PF 1SG:SBJ-16-Ieave here 'After getting problems, they wanted me to leave here.' The long final -áa does not seem to occur in Pare or Shambaa. The equivalent Pare form is SC-eki-stem-iye (Kotz 1909:13-14), in Shambaa SO/-stem-/7e, which is charac­ terised as resultative by Roehl (1911:142). Past imperfective ékú-: The past imperfective is most probably a fixed combination of ée PST and kú COND. The meaning is to indicate habitual past, imperfective past, and background in the past as in twéekudáha maéza, ... 'when we walked into the bush, ...' ékukwáa manú 'when they are two...'. The meaning of this "tense" contains something of both its constituting elements, past and conditional, but it is sufficiently different from these to warrant an analysis as a separate "tense" marker. Verb forms with this "tense" prefix take the default final vowel. This correlates with the incompatibility of the conditional with the suffix -iye — with the exception of verbs with a passive extension — and applies to both ku and ékú. Pare has a similar imperfective prefix eki (see Kotz 1909:8-9). Incidental: The incidental "tense" prefix is a high toned á, not to be confused with the falling tone áa in the past tenses. The prefix of the third person class 1 fuses with the "tense" marker, yielding á. This "tense" indicates punctuality of the action: 'at this point in time' ((48)) 'at a certain point in time', or 'at a conditioned point in time'. It is often used in the com­ plement phrase of a conditional phrase ((47)). It occurs only occasionally in my narra­ tive text materials, once for example in a short interruption directed at the interviewer ((48)). (5.47) ni-ku-só roró ná-kúru mbnvá nú(1Mb) lSG-COND-go 9:coast lSG:lNC-cultivate 9:field only 'If I go to the plains, I cultivate my field with effort.'

THE VERB

129

(5 48) w-á-síkia (NMb) 2SG-INC-listen 'Are you listening?' Optative H-_-...-e: The tone of the subject prefix in this "tense" is high which is the realisation of a preceding floating high tone. The third person singular class 1 concord is a and not e; the inflectional final vowel is e and no "tense" prefix is used. The high tone of the verb stem spreads to the final vowel e (if present), e.g., ádósé H-á-dós-e 3SG:follow 'Let him follow'. This "tense" expresses a wish on the part of the speaker. The same "tense" is attested in Pare. The "tense " prefixes te, ta, ne, na, re/ra, and ze/za: These four pairs of "tense" pre­ fixes have in common that they begin in an alveolar consonant and have a vowel that alternates between a and e. I will argue individually that te and ta are separate mor­ phemes, and similarly for ne and na, while in the other pairs a and e are allomorphs. These "tense" prefixes indicate primarily information structure and attitude of the speaker towards the state of affairs expressed in the clause rather than aspect or tense. Expectational ne: The "tense" prefix of the expectational is ne. Note that the "tense" marker na which is used in negative verb forms meaning 'not yet' is not an allomorph (see below). The first person singular concord ni is realised as n before ne, as it is in Pare with this same "tense" ne- (Kagaya 1989:32); Shambaa has two similar tenses: ne- near future and nee- distant future (Besha 1989:238). Although sentences containing this "tense" often refer to the future ((52)), it is also used for positive expectation ((49), (50)) or intention ((51)), and also for negative expectation ((53)). It is the expectation meaning that makes that this "tense" is obliga­ torily used in complements of an irrealis condition ((54)). (5.49)mirá va-gheni v-étu vé-ne-za saa vu-sikú (NMb) but 2-visitors 2-our 2-EXPEC-arrive hour 14.1-certain '(I go somewhere but) visitors should come at a certain time.' (5.50)u-na mburi ani hambu vi-nhtu vy-óse (NMb) 2SG-with 9:matter what because 8-things 8-all u-ne-vi-kicha 2SG-EXPEC-8-find 'What doubt do you have, because, all the things, you will find them.' (5.51)kwátó n-ne-endelea ku-bó'i na yá va-sítá na thus lSG-EXPEC-continue 15-make with this 2-Sh. with va- 'ariyé. (1Mb) 2-Pare 'Therefore I will (intend to) go on with these Shambaa and Pare.'

130

CHAPTER 5

(5.52)te-m-róána kúné nkaé  hódi m-né-gáya. (1Mb) NEG-2PL-leave:REC 2PL ago until place 2PL-EXPEC-die:APL'You don't leave each other from that time until you die.' (about marriage) (5.53) u-ku-sa-kom-wa, hé-ne-kom-wa úu 2SG-COND-NEG-kill-PAS 16-EXPEC-kill-PAS 1 :DEM 1 mwana wako lxhild l:your 'If you are not killed, this child of yours will be killed.' (5.54) í-ne-bigha hé í-li is anga vé-ku-ji-flnga j-ose 9-EXPEC-hit 16 5-5 5-land 2-COND-10-close 10-all 'It would rain in that land, if they closed it all.'

(NMb)

(NMb)

Counter-expectational na: The "tense" marker na is used in combination with the negative marker te, which is in pre-subject position, and a final high tone. It has the negative meaning 'not yet' ((55)). (5.55)vá-ngi vaana te-vé-na-ghuhwa-a (NMb) 2-other children NEG-2-EXPEC-marry:PAS-NEG 'Other children are not married yet.' One could imagine an analysis where ne 'expectational' and na 'counterexpectational' are contextually determined allomorphs since the meaning 'not yet' in negative "tenses" could be seen as an exponent of positive expectation (on the part of the speaker) with the negation markers expressing the negative part of the meaning 'not yet'. However, the use of na in affirmative sentences sheds another light on the meaning of this "tense" prefix because in those sentences it expresses indignation or counter-desiderative: compare the elicitated sentences ((56), (57). In texts there are instances where it indicates annoyance ((58)) or surprise ((59)). Thus while ne indi­ cates expectation and intention, na indicates the opposite, namely counter-expectation. (5.56) ni-ka-na-kúru (1Mb) 1 SG-CSEC-C.EXPEC-cultivate 'Why should I cultivate?! (indignantly)' (5.57) n-áa-na-kúru (1Mb) 1 SG-PST-C.EXPEC-cultivate 'I haven't cultivated yet! (indignantly)' (5.58)vé-ba-we m-ka-na-chele-ja 2-say-PAS:PF 2PL-CSEC-C.EXPEC-late-CS

wa tó sari saa (NMb) 9:cattle just until hour

131

THE VERB ikurni ni ahoni ten is what 'They were told: "Why do you make it so late for the cattle, until four o'clock?'" (5.59)niyo á-na-uma hé-dini kidogo (IMb) it:is 1-C.EXPEC-stand 16-religion little(Sw) 'He stopped with religion for a while.'

Involvement of the subject seems to be part of the meaning of both the ne and na "tense" markers. Consequently there is a high proportion of first and second person subjects used with ne or na in our texts. Swahili has a present tense marker na and thus na appears in code-switching as is the case in (60): the verb form tunasikia 'we hear' is Swahili, while the word imme­ diately following it, mazi, is Shambaa. (5.60)hambú áho bosi tunasihia mazi because here first we.are.listening(Sw) water(Sh) 'This is the first time we hear that the water is full.'

á-i-zua(NMb) 6-0.9-full

Logic sequential ta: The prefix ta (glossed with EVI) indicates that the action or event is according to what is to be expected, what should be, or what is a logical conse­ quence. While ne, as discussed above, assumes a possible influence of the subject on the event, ta excludes any influence of the subject. For this reason ne cannot be com­ bined with ta. The prefix ta is used, for example, in procedural texts ((61)) in historical ac­ counts indicating general custom ((62)), and in clauses supplementing a conditional ((63)). (5.6\)mhé é-zóko é'όro ndoá y-aki-Ma'á, é-ta-dáa na'á, man 1 -sit 1-close marriage 9-CON 7-Mbugu 1-EVI-seek honey é-ta-dáa churú, é-ta-dáa mipahé, é-ta-vá-se vahé, 1-EVl-seek bull 1-EVI-seek sugar.cane l-EVI-2-call people é-ta-m-sé mkóra kú'u w-a ki-'ágirú 1-EVI-1-call in-law his 1-CON 7-male na w-a kí-lagé na va-háli ú vée vábahú (IMb) and 1-CON 7-mother and 2-other they 2 siblings 'Somebody who wants to marry the Mbugu way looks for honey, a bull, sugar.cane and he calls people, he calls his in-laws of the male and female type and others, brothers and sisters.'

132

CHAPTER 5

(5.62) é-ta-tí'íwa ikeréno. ní-hó éhla y-etu (IMb) 3SG-EVI-carry-PAS earring is-16:REL custom 9-our 'He receives earrings. This is our custom.' (5.63 kai ni-ló baari, ni-ta-baari. (IMb) if 1SG-have journey 1SG-EVI-journey 'If I have a journey I'll travel.' Combinations of ta and other "tenses" occur in which ta is usually the last element, but not always: ta-za, ta-ranga, and ta-ku are all attested. In Pare ta is used to indicate that the speaker knows or assumes that the subject is involved in the action of the verb, cf. Kotz (1909: 24). Shambaa has the "tense" taindicating certainty, or decision (Roeh1 1911:113). The narrative "tense" te: The narrative "tense" prefix is te. This "tense" prefix allows the tone of the preceding subject prefix to spread onto it, and occasionally the tone spreading extends to the complete verb form, e.g., vé-té-ímá 2-NARR-cultivate (NMb). This "tense" is used to present the action or event as a logical sequence, with the meaning "it could not have been otherwise". It is particularly frequent in stories ((64)), and in historical accounts ((65)). (5.64)kwátó áni ni-te-daha anhao anhao m ni-ka-he maghamba thus I ISG-NARR-walk care care until 1 SG-CSEC-arrive M. 'Thus I walk on carefully until I arrive in Magamba.' (IMb) (5.65) bási vé-té-za-zéu sari tó hlenú. (IMb) well 2-NARR-IT-herd until there tomorrow 'Well they went to herd up to there the next day.' Although te and ta are proximal in meaning since they both indicate that what is ex­ pressed by the verb is "(self-)evident", they are different morphemes. The decisive argument against a mono-morphemic analysis comes from the fact that the combina­ tion of te and ta has been attested — be it only once ((66)). Elsewhere a sequence of the same "tense" prefix used twice is inadmissible in the verbal morphology. (5.66) tu-té-ta-lita (IMb) 1PL-NARR-EVI-come 'We should come.' Of the "tense" markers te and ta, ta is four times more common than te. The two forms are not found in strict complementary distribution but they would appear to be in a kind of statistical complementary distribution. Different types of texts favour ei­ ther te or ta; the former occurs more in stories and histories while the latter occurs

133

THE VERB

more in procedural texts. Both ta and te are used in Pare and in Shambaa. In both languages the two forms are analysed as being different although they have similar meanings. Pare has a "tense" form te- meaning 'really' (Kotz 1909:25); in Shambaa, there is also a "tense" form te- indicating that something is self-evident, or logical, or remarkable on the ba­ sis of the action or event, see Roehl (1911:113). Background re/ra: The background "tense" prefix is re or ra. I consider them to be variants. Both re and ra occur in a very similar context, namely after ka-kwa 'and-is' ((67), (68)). (5.67)mirá té-hlamáye a-ka-kwá é-re-úma sawasawa bi. (IMb) but NEG:l-leave:NEG 1-CSEC-be 1-BGND-stand equal only 'He did not end up staying in a standstill.' (5.68)vé-ka-buu vé-ka-kwa vé-ra-salia i'i (IMb) 2-CSEC-enter 2-CSEC-be 2-BGND-pray here 'In the end they entered with the children, they entered and prayed here.' It is not clear whether the subject prefix has the default vowel e in this "tense" or the vowel a instead. In actual texts in IMb and NMb the vowel of the subject prefix is just as often a as e. In elicitation, however, my main informant for Inner Mbugu did not have a in the subject prefixes, but only e, while my main informant for Normal Mbugu had both a and e. The variant re occurs more often than ra in my collection of texts. It is used to indicate something that is not in the natural flow of events; or to indicate for example a habit (in the past) against which background the action has to be interpreted, or a general background against which the action is narrated; it is often translated by 'when' (as in (69)) and I gloss it as a background "tense" BGND. (5.69)ání ni-aho-íye tó aba é-re-ata i'i I 1SG-see-PF just father 1-BGND-be here 'I saw it when father was still here.'

(IMb)

The background "tense" is also used to place events before the point of reference on the story line. The sentence á-ré13-bátijwa péré 'he was baptised already' refers to an event that had happened in the years before the reference time at that point in time of the story. A few sentences later we find a similar example of re used for a previous event that is now relevant background ((70)). (5.70)kå padre a-re-gá niho kenedi é-lita (IMb) DEM2 priest 1-BGND-die is-16:REL Kennedy 1-come 'After that priest had died, father Kennedy came.'

134

CHAPTER 5

The label "background" certainly does not cover every usage of the marker; there is one attestation of ra being used to indicate an obligation ((71)). Nor does it account for the fact that this "tense" is often used with recipient telic verbs such as 'get to know', (72), 'die', 'slim', 'be named'. (5.71)kai ni ŋornbe m-ra-ghava sawasawa. if COP cow 2PL-BGND-divideequally(Sw) 'If it is cows, you are to distribute equally.'

(NMb)

(5.72)jo

(NMb)

10:REL

ni-re-manya.

jo

u-re-manya.

1SG-BGND-know 10:REL 2SG-BGND-know

'Those I know? Those you know.' In the form re the "tense" marker is identical to one of verbs 'to be'. However, while the verb 'to be' has an allomorph nde (see section 5.5); the "tense" marker re has no such allomorph; inversely, the verb 'to be' does not have a variant ra. Historically the origin of the "tense" marker might well lie in the verb 'to be'; compare the parallel in English usages such as being for background, and is to for obligation. In Pare both forms re and ra occur but whether they are variants is not com­ pletely clear from the literature on Pare. The re- form is used as a subsequent subjunc­ tive according to Kotz (1909:15) and the ra- form "signals simultaneous reference to more than one time" according to Mreta (1998:96). In the perfect, the relation between the two appears: ra- is replaced by re-, cf. Kotz (1909:18,19). Kagaya (1989:35) gives both forms under the label Present Continuous and remarks that re- also expresses 'the subject's intention or schedule at the moment'. Shambaa seems to lack re- and ra-. Itive za/ze: The movement marker za or ze occur before or after the object prefix; in texts the itive markers always preceded the object prefix. The subject prefix normally contains an e, but occasionally a. Both forms of the prefix are translation equivalents of the Swahili consecutive prefix ka. For example ázaima or ázeíma 3SG:IT:cultivate (NMb) 'and he culti­ vated/cultivates, he goes/went to cultivate' is rendered in Swahili as akalima and ázepáta 3SG:IT:get (NMb) 'and he got/gets' is in Swahili akapata. This in itself is an indication for the identity in meaning of ze and za and thus for the allomorphy. The allomorph za is more common than ze; in several instances we have noted that they are interchangeable. The meaning of the "tense" marker is itive, i.e. 'go and ...' ((73), (74)); as an ex­ tension of this meaning the marker is also used for (past) consecutive, (75), and near future (without necessarily involving movement ((76)); it is also used to indicate in­ choative meaning, i.e. entry into a state ((77)). (5.73) hé-ló i-'azé 16-have 5-day

i-wé áa-sé va-Ma' á na va-sitá na 5-one l:PF-call 2-Mbugu and 2-Shambaa and

135

THE VERB

va-'ariyé vá-so vá-zé-m-hand-íya ma-gerú kú'u. (IMb) 2-Pare 2:SBJ-go 2:SBJ-IT-l-plant-APL 6-bananas his 'On a certain day he called the Mbugu, Shambaa, and Pare people to go and plant his banana trees.' (5.74)é-tonga é-za-mu-kela 1-walk 1-IT-1-explain 'He went to explain to him.' (5.75)vé-za-tagh-wa hé 2-IT-sell-PAS 16 2-arabs 'And they were sold to the Arabs.'

(NMb)

va-arabu

(5.76) tú-zé-ga'á-we msíndo na hé-ruvú. lPL-IT-kill-PAS:PF bang with 16-Ruvu 'We are going to be killed, wham, into the Ruvu River.'

(NMb)

(IMb)

(5.77)e-ká-itang-wa mwambéni m-nhkangwaná nï-re-yé m-lóndo 1-CSEC-call-PAS Mwambeni 1-Nhkangwana is-BGND-s/he 1-1. a-za-ikáa hé lu-ngorí lw-á wa-nhkangwána wakatí uo; 1-IT-stay 16 11-hill 11-con 2-Nhk. 3:time 3:DEM2 íngeí vá-la va-nhtu nhkondo ji-ne-kúnda now 2-DEM3 2-persons 10:war 10-EXPEC-want ku-za-ku-dósa-ir-ij-a he-ti-wa he- i-tawi 15-IT-COND-follow-APL-CS 16-say-PAS 16-with 5-branch l-aa-bajika uko Vudee l-a va-mbughu. (NMb) 5-PST-divide there Vudee 5-CON 2-Mbugu 'He was called Mwambeni, a Nhkangwana, that is a Londo, and he stayed at the mountain of the Nhkangwana that time; now those people wanted war to follow them; there was a branching that divided there at Vudee, the branch of Mbugu.' The itive prefix often follows other "tense" prefixes. Attested combinations include aa/ee-za, ta/te-za, ne-za, ku-za. The prefix za is always the last in such combinations,14 and when combined with other "tenses" it never takes the form ze. The origin of the prefix is most probably the verb za (NMb) 'to arrive' (cf. the imperfective conditional he below). This origin is still reflected in the fact that the pre­ fix is frequent with ventive verbs such as 'come', 'bring', 'arrive', 'enter', 'return'. Pare has a "tense" marker (z)e- meaning 'in order to' (Kotz 1909:16). Also in Mbugu the z of the itive marker is sometimes dropped. Consecutive ka: In meaning, the prefix za/ze comes very close to the consecu­ tive marker ka; however, the two prefixes can co-occur, always in the order ka-za. As a "tense" marker on its own, ka with consecutive meaning is more common; as an ad-

136

CHAPTER 5

ditional "tense" marker ze/za is more common. I will call the ka tense consecutive in order to differentiate it from the itive. The prefix ka is often used as only "tense" pre­ fix, more so than za/ze, and then it refers to consecutive and also to future action. Shambaa has ka- 'now and again' (Roehl 1911:116). Pare (Kotz, 1909:23) and Swahili have ka- as consecutive or narrative. Conditional ku: The conditional "tense" marker is ku. It causes the tone on the preced­ ing syllable, the subject prefix, to become high. A floating high tone in the position preceding the subject prefix is assumed. There is considerable tonal variation in this "tense"; in many instances ku is not high and there is no high tone mapped onto the subject prefix as in example tukuvóna in the list. The conditions for this behaviour are not clear. The "tense" prefix ku is sometimes replaced by its Swahili equivalent ki. The prefix ku can be combined with many others, in which case it adds "certainty", com­ pare hénézakuhláwa 'there will be suffering for sure' with hénézahláwa 'there will be suffering'. Pare has no conditional "tense" prefix; instead it uses an adverbial and a verb form with eki- (Kotz 1909:23); Shambaa as well as Swahili express conditional mean­ ing with a prefix ki- (Roehl 1911:115). Thus Bantu languages in the area all seem to have ki where Mbugu has ku for this "tense". I have no explanation for this anomaly. Imperfective conditional he: Just like the preceding conditional "tense", the imperfective conditional "tense" has a high tone on the subject prefix, due to a floating high tone in pre-subject prefix position. The subject prefix has the a variant and not the usual e variant in the subject prefix. This "tense" is used for (past) imperfective situa­ tions that are the condition or circumstances for the following phrase ((78), (79)). All my examples with he are from Inner Mbugu, but this may also be due to chance as the total number of examples in my texts is very low. The historical source of the marker he is probably the agreement prefix of locative class 16 he. Pare does not seem to have a "tense" marker he-, although Shambaa does, meaning 'just', (Roehl 1911:114). (5.78)Maganga á-he-líta sé-áta Maganga 1-IMPF.COND-come 1SG:NEG:PST-be 'When Maganga came, I wasn't there.'

(IMb)

(5.79)n-he-wesía-wa, ni-áho é-té-ghesha hú shuhé (IMb) 1 SG-IMPF.COND-watch-PS 1 SG-see 1 -EVI-try that leopard 'While I am being looked at, I see that he measures, that leopard.' The polysyllabic "tense" markers tuwáá/túwa, chéri, renge/ronge, and mangá are re­ cent intrusions into the "tense" marking system. They can be followed by other "tense" prefixes such as ku, he, za/ze. Two of them, chéri and túwa cannot precede imperatives. These four "tense" prefixes do not occur independently as adverbs other-

THE VERB

137

wise. These polysyllabic "tense" prefixes are not reported for Pare or Shambaa. Habitual tuwáá/túwa: The "tense" marker is tuwáá or túwa with variation in the tone and vowel length. The long final vowel was only recorded in the Magamba-area. The meaning of túwa/tuwáa is habitual, irrespective of tense, past or present. This "tense" marker was not attested in texts. Present unfinished (still) chéri: The present unfinished "tense" marker is chéri and its meaning is to emphasise that the action is still going on. Hortative réngé: The hortative "tense" marker is réngé, with all sorts of variations such as rengé, ronge, ranga, ronga (glossed with FIRST). The origin of this marker, which is common to both Inner and Normal Mbugu, is the Normal Mbugu verb ronga 'to do, make'. The variation is possibly inspired by a reanalysis of the first syllable of the prefix as the "tense" prefix re/ra. The prefix expresses a hortative meaning, e.g. túréngé-héma 1PL-FIRST-rest 'let us rest first.' (IMb). Another aspect of the meaning is to indicate that the action is only a beginning, "first" as in in niréngékúru 'I cultivated here first'. Note that it is a Normal Mbugu lexical form that gets grammaticalised; grammaticalisation from Inner Mbugu lexicon does not occur at all. Pare has the same "tense" prefix ronga (Mreta 1998:64). Speed mangá: The "tense" marker mangá adds a sense of hurry, speed, e.g., é-mangákúru 1-SPEED-cultivate 'he hurries to cultivate' (IMb). In imperatives it indicates immediateness ((80)). (5.80)mangá só mangá litá mangá tí'i 5.3

'Leave immediately!' (IMb) 'Come immediately!' (IMb) 'Lift it immediately!' (IMb)

Combinations of"tense"prefixes Combinations of "tense" markers are possible; especially frequent are combina­ tions with conditional ku followed by another "tense" prefix, and the markers re/ra or te/ta following either the past marker ée/áa or the consecutive ka. Only those combi­ nations that have actually been attested in texts are mentioned below; not mentioned are the very frequent combinations of the itive ze/za with other "tense" prefixes. There are probably only semantic limitations to the combination of "tense" prefixes and some combinations that did not occur in texts were actually accepted in elicitation, such as kú-ne and ku-tuwa in ni-kú-ne-ima (NMb) and ni-ku-tuwa-ima (NMb) 'when I cultivated'.15 A combination of three "tense" prefixes was also accepted: n-ée-te-reíma 1SG-PST-EVI-BGND-cultivate 'when I was busy cultivating' (NMb). The exact se­ mantics of these combinations are difficult to establish as the number of tokens in texts is low (the combinations ka-ra CSEC-BGND (81) and ékú-ta PST.IMPF-EVI ((82)) both occurred only once) and the meanings are often very near to those of one of the

138

CHAPTER 5

two (or more) tense prefixes. Translation equivalents in Swahili are unreliable when establishing precise nuances of meaning. For example the combination ée-te PST-EVI refers to a general past action in (83) and ta-ranga EVI-FIRST does not differ much from ranga FIRST in (84). (5.81)na-yé é-ku-za a-ka-ra-ßv-í-w-a. with-1 :REL 1 -COND-arrive 1 -CSEC-BGND-die-APL-PAS-F 'And when she arrived she died (she was met with death). '

(NMb)

(5.82)na ku-hlamí-w-e tw-éeku-ta-butú-w-e makeréno. with 15-anoint-PAS-PF 1PL-PST.IMPF-EVI-put:in-PAS-PF earrings 'And, anointed, we were put on earrings.'

(IMb)

(5.83) toká m'azé ki-vóko ghó, n-ée-te-ómba m-zúngu mu-wé since day 7-start my 1SG-PST-EVI-request 1-european 1-one é-sé-wé maja sandres (IMb) l-call-PAS:PF major S 'Since the day of my beginning, I have requested a European, called Major Sanders.' (5.84)n-ta-ranga-ro 1 SG-EVI-FIRST-leave now 'I leave it first.'

iji.

(IMb)

Examples in elicitation and texts sometimes point to different meanings, for example, in the case of the combination ku-ta COND-EVI, which occurred with a simple narrative meaning in a text about the initiation period ((85)); in elicitation, however, various Swahili renderings were given which all stay closer to a conditional sense: "endapo nikilima", "nilikuwa ninalima", or "nikilima" for the forms ni-kú-ta-kúru (IMb) or nikú-ta-íma (NMb) 1SGxultivate. This was also the case with other Swahili renderings of combinations involving ku, see above. (5.85)iji aré roró ni-ku-tá-bísh-w-a bí. now there coast 1 SG-COND-EVI-pester-PAS-F only 'There at the plains I was pestered all the time.'

(IMb)

Combinations of "tense" prefixes can acquire new meanings which is most probably the history of the "tense" marker ékú. The combination kú-ra/re COND-BGND may very well be in a similar process since it seems to refer to a single occasion as in u-kurasíkia 2SG-EVER-hear 'Have you ever heard of, or (86).

THE VERB (5.86) u-kúre-mana na mnhtu, 2SG-EVER-m¡ss with person 'If you don't have anybody,'

139 (NMb)

Another case where the meaning of the prefix combination is not the simple sum of the combining markers is ka-ta CSEC-EVI, which in all tokens expresses a condition, uka-ta-vóna 2SG-CSEC-EVI-see 'if you see' (as in (87), (88)'and (89)). Sentence (87) is a continuation of (86). (5.87) u-ka-ta-chi-m-lót-ij-a uvu, é-ta-fwá. 2SG-CSEC-EVI-7-1-point 13:DEMl 1-EVI-die 'If you point him thus, he will die.'

(NMb)

(5.$$)ni-hó e-sí-ka-tóng-e tu-ka-ta-m-shigh-ir- ij-a ko is-16.-REL 1 -NEG: SBJ-CSEC-gO-NEG 1PL-CSEC-EVI- -let-APL-CS (NMb) mkalá. hunter 'So that he doesn't leave, if we leave him, that hunter.' (5.89)vé-ti-wa ará mbwánji m-ka-tá-vóna va-nhtu ve-téta ki-zíghúla. 2-say-PAS ara sorry 2PL-CSEC-EVI-see 2-persons 2-talk 7-Zigua 'They were told: If you see people talking Zigua, ...' (NMb) In addition to the combination of "tense" prefixes within one verb form, combinations are also possible by using two verb forms, the first one being the verb "to be". Fur­ thermore, verb forms may be preceded by a copula with a relative locative pronoun, yielding nihó, e.g., adding an epistemic value of certainty (see section 5.5). 5.4

The negative "tenses" Negation is maximally expressed in three positions in the verb: the presence of a negative prefix preceding the subject prefix, the presence and form of a negative fi­ nal inflectional suffix, and the (negative) "tense" prefix. The options are summarised in Table 5.8. negative prefix te, Ø final suffix é, H, PF, Ø negative tense prefix si Table 5.8: Expressions of negation in verbs Most negative "tenses" have a prefix te preceding the subject prefix and a high tone on the following syllable, that is, on the subject prefix. The same high tone is used in the optative and conditional and we take "non-assertive mood" as the function of this

140

CHAPTER 5

floating high tone. The prohibitive and the negative conditional are the only negative "tenses" that lack this initial prefix te, but they do have the initial high tone of nonascertainment. The meaning is prohibitive rather than denial of a wish, and the condi­ tion of a situation expressed by a negative rather than the absence of a condition. Thus, for these two "tenses", the negation is in the (logical) dependent domain, which is different from the other negative tenses and probably the reason why they lack the initial negative marker. Some negative "tenses", primarily those that contain a negative "tense" marker, have a negative final element -é or a final high tone.16 The verbs in Inner Mbugu that do not take the inflectional final element including those that end in an invariable final a such as waha 'drink', do not take these final elements either, thus u-sí-ká-wáha 2SG-NEG-CSEC-drink 'you are not to go to drink'. The perfect suffix -íye does not cooccur with a negative inflectional suffix. There are a number of "tense" markers that are limited to negative "tenses" or that have a specific function in negative "tenses". These are: ka, na, hå, si, sée, sa and the combined "tense" prefixes kahá, kúséé. The other "tense" prefixes can be used in both affirmative and negative "tenses", and may also be used in addition to these negative-"tense" prefixes. In negative "tenses", as is the case in affirmative "tenses", "tense" prefixes can form a series, e.g., ni-si-te-za-ghoja 1SG-NEG-EVI-IT-wait 'I am not going to wait for it' (NMb). Some "tense" markers and some allomorphs of "tense" markers do not occur in negative "tenses", namely: áa PST; áa..á COMPL; ra BGND; ne EXPEC, and perhaps even some others. There is one negative non-past marker ka which is homophonous with the con­ secutive marker ka. The two can never be confused even though the consecutive may occur in a negative "tense". The consecutive ka can only occur in a negative verb form if it follows the negative "tense" prefix si and in that position the negative non-past ka cannot occur. Examples are: e-si-ka-tong-é 1-NEG-CSEC-walk-NEG 'he is not to go for a walk', u-sí-ká-wáha 2SG-NEG-CSEC-drink 'you are not to go to drink', and tu-sá-ka-jíle 1PL-NEG-CSEC-10-know 'we won't know them'. This usage is parallel to the itive as in m-si-ze-ikaa 2PL-NEG-IT-stay 'Don't stay!' Some situations that contain a negative in an English rendering are described by an affirmative verb form with a "negative" verb, as in tw-áa-ro ku-gá'a 1PL-PST-leave 15-kill 'we stopped killing'. Given the mismatch between the "tense" prefixes that occur in affirmative and those that occur in negative "tenses", and given the fact that there are nega­ tive/prohibitive prefixes that take the position of "tense" prefixes, it becomes evident that the negative "tenses" cannot be considered as simple transformations of affirma­ tive "tenses". Thus, it is better to consider the negative "tenses" as "tenses" in their own right. The main division among the "tense" prefixes is between those that are inher­ ently negative or prohibitive, such as si, and the others. For those "tense" prefixes that are negative/prohibitive, there is no negative pre-subject prefix te. When the negative

141

THE VERB

prefix te is present, the set of possible "tense" prefixes is not completely identical to the one used in affirmative "tenses", that is, some of the "tense" prefixes may only occur in a negative verb form. Given these and other restrictions on combinatory pos­ sibilities of the variables of Table 5.8 above, all negative "tenses" can be summarised in Table 5.9 and Table 5.10. The "tense" markers in italics occur in negative "tenses" only. final suffix Ø

attested "tense" prefixes Ø, ta, re/ra, ze, tuwaa, cheri, renge, eku, sée, sa PF 0, eku, sée, há H na 0 , ka, kahá é Table 5.9: Overview of verb forms with a negative pre-subject prefix te final suffix negative "tense" prefixes 0 si, kusa, se e si, kúséé Table 5.10: Overview of verb forms with a negative "tense "prefix sV Examples for these negative "tenses" in the following Table 5.11 are given according to the arrangement in Table 5.6. But I used more different verbs and there are more gaps in the data. IMb general negative te-tú-kúru te-vé-kúru te-tú-doa te-vé-doa

GLOSS

NMb

GLOSS

Ø NEG-1PL-dig NEG-2-dig NEG-1PL-touch NEG-2-touch

negative logical sequential

negative baclground sí-re-kúru

NEG: 1 SG-BGND-dig

sí-re-doa

NEG: 1 SG-BGND-touch

te-vé-ima

NEG-2-dig

ta

te-tú-ta-ima te-vé-ta-fínga

NEG-1PL-EVI-dig NEG-2-EVI-close

re/ra te-vé-ra-dósa

NEG-2-BGND-follow

142

CHAPTER 5

negative itive sí-ze-kúru

NEG:lSG-IT-dig

sí-ze-doa

NEG:1SG-IT-touch

NEG:l-HAB-dig

NEG:l-HAB-touch té-túwáa-doa negative present unfinished téchérikúru

réngé, ronga

NEG:l-FIRST-dig

NEG:l-FIRST-touch té-réngé-doa negative imperfective past te-v-éku-dúmu

cheri

NEG:l-STILL-dig

NEG:l-STILL-touch techéridoa negative past unfinished té-réngé-kúru

NEG:lSG-IT-eat NEG-2-IT-eat (NMb)

túwáá

negative habitual té-túwáa-kúru

ze sí-zé-lá te-vé-zé-lá

te-vé-ronga-ima éku

NEG-2-FIRST-dig

te-v-éku-vika

NEG-2-IMPF.PST-put

NEG-2-IMPF.PST-want

negative irrealis

sée

te-tú-sée-kúru te-vé-sée-kúru

NEG-1PL-IRR-dig NEG-2-IRR-dig

negative perfect te-tú-kúrúye te-vé-kúrúye te-tú-do-íye te-vé-do-íye

NEG-1PL-dig:PF NEG-2-dig:PF NEG-1PL-touch-PF NEG-2-touch-PF

negative past perfect NEG-lPL-PST-dig:PF te-tw-ée-kúrúye NEG-2-PST-dig:PF te-v-ée-kúrúye NEG-1 PL-PST-touch-PF te-tw-ée-do-íye NEG-2:PST-touch-PF te-vée-do-íye

te-tw-é-sée-vóna te-sée-vóna

NEG-1PL-PST-IRR-see NEG:l-IRR-see

Ø sí-ang-íye

NEG:1SG-walk-PF

té-vé-im-íye

NEG-2-dig-PF

ée

143

THE VERB

há té-há-kúrúye

N E G : 1 -IMPF.COND-dig:PF

té-há-do-íye sée

te-tú-sée-do-íye te-vé-sée-do-íye not-yet te-tú-na-kúru te-vé-na-kúru te-tú-na-doá te-vé-na-doá

NEG-1 PL-NEG-touch-PF

na NEG-2-C.EXPEC-dig:NEG NEG-lPL-C.EXPEC-touch:NEG NEG-2-C.EXPEC-touch:NEG

te-vé-na-imá

NEG-2-C.EXPEC-dig:NEG

0

sí-kuru

NEG:1SG-dig:NEG

sí-do-é

NEG:lSG-touch-NEG

ka NEG-1 PL-NEG.FUT-dig:NEG NEG-2-NEG.FUT-dig:NEG NEG-1 PL-NEG.FUT-touch-NEG NEG-2-NEG.FUT-touch-NEG

negative conditional NEG:l-NEG.COND -dig:NEG

té-kahá-do-é

NEG:l-NEG.COND -touch-NEG

negative relative e-sí-áxo

si-ka-í-é té-ka-dós-é te-tú-ka-im-é te-vé-ka-im-é

N E G : 1 SG-NEG.FUT-eat - N E G NEG: 1 -NEG.FUT-follow-NEG NEG-1 PL-NEG.FUT-dig-NEG NEG-2-NEG.FUT-dig-NEG

kahá

té-kahá-kúru

sí 1-NEG-see

e-sí-idima negative conditional

1 -NEG-can

kusa u-kusa-síkia e-kusa-komwa

prohibitive tú-si-kúru vá-si-kúru tú-si-do-é

NEG: 1 SG-NEG-5-dig-PF

NEG-lPL-C.EXPEC-dig:NEG

negative incidental

negative non-past te-tú-ka-kúru te-vé-ka-kúru te-tú-ka-do-é te-vé-ka-do-é

si-séé-li-im-íye

NEG-2-NEG-touch-PF

si lPL-NEG-dig:NEG 2-NEG-dig:NEG 1PL-NEG-touch:NEG

2SG-NEG.COND-hear 1-NEG.COND-kill

144

CHAPTER 5

2-NEG-touch:NEG vá-si-do-é negative conditional NEG-1 PL-C.EXPEC-dig:NEG tú-kiiséé-kúru é-kúséé-kúru NEG-1 -C.EXPEC-dig:NEG tú-kúséé-doá NEG-lPL-C.EXPEC-tOUCh:NEG vé-kúséé-doá NEG-2-C.EXPEC-tOUCh:NEG

vá-si-im-é kúséé

2-NEG-dig:NEG

Table 5.11: The negative tenses with a negative pre-subject prefix te Remarks on the negative tenses: Negative itive: No za is attested as first "tense" prefix in a negative verb form. The tonal assimilation of the "tense" prefix is optional. Negative Present Unfinished: For the negation of an affirmative verb form with cheri, either this form is used or the negative non-past. Negative imperfective past: The translations of these forms are 'they didn't want' and 'he didn't put', respectively. Negative irreal is: The prefix can be preceded by the past "tense" prefix é with no ap­ parent additional meaning, té-mw-e-sée-vóna NEG-2PL-PST-IRR-see (NMb) 'if only you had not seen, you should not have seen'. The translation of these negative forms is the Swahili singe as in ni-singe-ona. Negative with tense prefix sa: I have only one example of this "tense" prefix which contains other inexplicable elements such as chowe: t-u-sa-cho-we-yo NEG-2SG-NEG7:REL-one-say 'Don't say a thing!' Negative incidental: The "tense" can be used as denial of the habitual. Negative non-past: The tone of the negative final is realised on a preceding applied suffix making it homophonous with the perfect suffix, (90). (5.90) te-tú-ka-va-ba-í-e kigi NEG-1PL-NEG.FUT-2-tell-APL:NEG thing 'We won't tell them anything.'

(IMb)

Negative relative: Examples of the use of this "tense" are: esímurú kuyó 'one who cannot speak', ésikóíya 'one who does not hear', esíáxo 'one who does not see' (IMb). There are some prohibitive verb forms that lack the negative final é and are thus for­ mally identical to the negative relative although in meaning they are clearly prohibi­ tive ((91)); all four such examples that I have are from Normal Mbugu and the final vowel is of the inflectional type.

145

THE VERB (5.9)u-si-finga luvwi vé-si-pungua m-si-ze-ikaa ni-si-té-za-ghoja

2SG-NEG-close door 2-NEG-diminish 2PL-NEG-IT-stay 1 SG-NEG-EVI-IT-wait

'Don't close the door'. 'They are not to get less. 'Don't stay' 'I am not going to wait for it'

Negative Conditional kusa: Examples of the meaning of this "tense" are (92) and (93). (5.92) ekusakómwa 1-NEG.COND-killed 'if you are not killed*

(NMb)

(5.93) u-kúsa-ni-vwira níni 2SG-NEG.COND-1SG-say what 'why didn't you tell me?'

(NMb).

Negative past with se: The only two attestations of the negative past are (94) and (95) from Normal Mbugu.

5.5

(5.94) á-se-kéta 1-NEG-take out 'He didn't take out'

(NMb)

(5.95) a-se-finga 1-NEG-close 'He didn't close'.

(NMb)

Copula and verbs "to be" Sentences that contain no verb, contain either a copula ni or a construction con­ sisting of a subject concord plus the preposition na 'with, and'. Both constructions can be tensed. When tensed or when a subject concord is present, a copula construction uses the stem re. In addition, there are verbs in both Inner and Normal Mbugu with no substantial lexical content, i.e. verbs "to be" and verbs with a locative meaning "to be present, at a certain place". The two languages are not completely parallel in this area of syntax. The construction with na expressing, among other things, possession, is limited to Normal Mbugu; to express possession, Inner Mbugu has a verb "to have", lo. Both languages use the same copula ni in the same way including its replacement re. The locative verbs "to be" are not completely parallel in usage: the Normal Mbugu locative áho is much more limited in use when compared to its Inner Mbugu equiva­ lent áta. The copula and the verbs 'to be' and their functions are summarised in Table 5.12 below.

146

form ni si re SC+na áta áho kwá~xwá óka

CHAPTER 5

function or type of complement tense prefixes equation no negative equation no equation, insistence, emphasis yes possession, association, presence yes locative, presence, various complements yes locative yes qualitative, numeral, existence yes qualitative, compound "tenses", numeral, yes existence, compound "tenses" Table 5.12: Copula and verbs "to be " and their functions

language I+NMb I+NMb I+NMb NMb IMb NMb IMb NMb

The copula ni and its negative equivalent si are used in both Normal and Inner Mbugu with nominal and adjectival complements, such as a noun ((96), (97)), an independent demonstrative ((98), (99)), a nominal phrase ((100)), an adjective ((101)), or a preposi­ tional phrase ((102)). All examples are specificational and not predicational in the sense that Declerck (1988) defines this distinction. (5.96) kilúgwi ni kinyongôlé chameleon COP insect:Q 'Is the chameleon an insect?'

(IMb)

(5.97) ihohó si himéno bat NEG:COP bird(s) 'The bat is not a bird.'

(IMb)

(5.98) iduka li-hlá-w-a ni hú shop 5-open-PAS-F COP DEM3 'The shop that is open is that one.'

(IMb)

(5.99) ní ú é-mw-omb-íye COP l:DEM1 1-1-request-PF 'It is he who requested him.'

(NMb)

(5.100)

kigí yá si cha thing this NEG:COP 7:CON 'This thing is not copper.'

(5.101) mshwá ni m-kusá ant COP 3-bad 'White ants are bad.'

shaba copper

(IMb)

(IMb)

147

THE VERB (5.102)ni sa vigí, ni makeréno, COP as things COP earrings 'It is like things, like earrings.'

makeréno earrings

(IMb)

The juxtaposition of two nominal phrases is not enough for a predication; a copula is needed, as is evidenced by the fact that a translation of the Swahili sentence mti mwebamba 'tree (is) thin' (which contains no copula) will automatically have a copula added to it ((103)). (5.103)

mxatú ni m-chachó tree COP 3-thin 'The tree is thin'

(IMb)

When a relative pronoun of the locative class 16 is suffixed to the copula, yielding niho COP-16:REL, or sí-hó NEG:COP-16:REL, a verb form can follow this complex ((104),(105),(106)). The copula complex has the epistemic quality of asserting what follows. In stories, and especially in historical accounts, it acts as a linking device be­ tween the narrative steps. (5.104) nihó vevóka ku-kwá kanísa COP-16:REL 2-begin 15-build church here 'And then they started to build a church here.'

i'i

(IMb)

(5.105) ni-hó hanhtu v-áa-fumáa, vudée nihó (NMb) COP-16:REL place 2-PST-come:from-PST Vudee COP-16:REL mtasó w-avo prayer 3-their 'It is the place they came from, Vudee is their prayer.' (5.106) sí-hó i-líijanya (IMb) NEG:COP-16:REL 9-put:out 'It is not that it came out.' The construction subject concord plus na 'be with' is used for possession ((107)), as­ sociation ((108)) and even location ((109)) in Normal Mbugu. The construction can be conjugated for person ((110)), tense ((111)) and polarity ((112)) as if it were a verb. The "tense" prefixes te NARR and ée PST have been attested. (5.107) háta m-te-ná mahémba (NMb) even 2PL-EVI-with maize 'Even if you have maize.'

148

CHAPTER 5

(5.108) lukándo lu-na luzíghi wall 11-with rope 'The wall has  .'

(NMb)

(5.109) ndighe jí-te-ná mtí grasshoppers 10-EVI-with tree 'The grasshoppers are in the tree.'

(NMb)

(5.110) tw-ée-na, mw-ée-na (na) mafuta 1PL-PST-with 2PL-PST-with (with) oil 'We had, you had fat.'

(NMb)

(5.111)ili isanga te-l-ée-na mwasú 5:DEM1 land NEG-5-PST-with Pare 'This land had no Pare.'

(NMb)

The complex is sometimes followed by na again, as in ((110)) above and in ((112)). A further example is ((116)) below. (5A\2) vé-na na magúro 2-with with dogs 'They had dogs.'

(NMb)

The construction is often used with a locative subject concord, he-, to express pres­ ence ((113), (114), or its negative equivalent te-he-na for absence ((115)); and these constructions too can be tensed by ée PST or by te NARR. The complement of na may be fronted (see (115), (116)). (5.113) ká hé-na mghení home 16-with visitor 'At home there is a visitor.'

(NMb)

(5.114) héna mahírímo 16-with changes 'There are changes.'

(NMb)

(5.115) iyó majumbí te-hé-na today clouds NEG-16-with 'Today there are no clouds.'

(NMb)

(5.116) hambúi jí-la bahári vée-na najó há-la (NMb) because 10-DEM3 sea 2:PST-with with-10:REL 16-DEM3 'Because of those seas that they have there.'

THE VERB

149

The construction described above is common in Bantu languages; Inner Mbugu, how­ ever, has a lexical verb 'to have, possess', lo and is in this respect more similar to a Cushitic language such as Iraqw, that also uses a lexical verb 'to have'. The verb lo in Inner Mbugu has, however, also been recorded with the general purpose preposition na following it ((118)), compare also (117) with NMb (110) above. While Inner Mbugu lo 'to have' and Normal Mbugu "be with" may seem to be very different struc­ turally, in certain respects they look very similar: Both have possession expressed in a prepositional phrase with na 'with'; both take a class 16 locative prefix to express the presence of existential meaning ((119)), and for the existential presentative construc­ tion that occurs at the beginning of a story ((120)). (5.117) tw-ée-ló na mahlóma 1PL-PST-have with fat 'We had fat.'

(IMb)

(5.118) mw-a-li va-kabúna háta va'inyí vó m-ló na ú 2PL-INC-come:from 2-all even children they 2PL-have with 1 'You come all, with the children that you have,' (IMb) (5.119) hé-ló mwagirú é-sé-we kimwérí l-call-PAS:PF Kimweri 16-have elder 'There once was an elder called Kimweri.'

(IMb)

(5.120) hé-ló i'azé i-wé 16-have day 5-one 'On a certain day (once upon a time)'

(IMb)

The general preposition na has also been reported with verbs "to be" such as re-na ((122), (123)), ata-na, and oka-na. The complex SC-"tense" prefix-re is used with nominal complements ((121)), or in the "be with" construction if tense or a (first or second) person subject has to be expressed (see example (122)). The form of re in the first person singular is nde ((123)). (5.121) tw-ée-re kinhtu íchi ki-mwe 1PL-PST-BE thing 7:DEMl 7-one 'We are one.'

(NMb)

(5.122) na íki ni-ré-na na-yé with now 1SG-BE-with with-l:REL 'And I have been with him up until now.'

(NMb)

150

CHAPTER 5

(5.123) áva vá-ngi v-óse ndé-na na-vó 2:DEMl 2-other 2-all lSG:BE-with with-2:REL 'Those others, I have them all.'

(NMb)

This complex with re can still be preceded by the copula and such a construction is used to express emphasis. The complement is either a noun ((124)), or a conjugated verb form ((125)). The complex can be followed again by a relative pronoun agreeing with the subject ((126), (127)), or a locative relative pronoun ((128)). (5.124) kazi y-avó n-ée-re u-wínda-winda dú work(Sw) 9-their COP-PST-BE 3-hunt(Sw)-hunt(Sw) only 'Their work was just to hunt.' (5.125) vatúmwa n-ée-re ve-ra-víkwa vatúmwa slaves COP-PST-BE 2-BGND-put:PAS slaves 'The slaves were put with the Arabs.'

(NMb)

na mwarábu with Arab (NMb)

(5.126) va-mbughu n-ée-ré-vó va-nango 2-Mbugu COP-PST-BE-2:REL 2-Nango 'The Mbugu are Nango (Shambaa people -Mbugu descent)'

(NMb)

(5.127) inŋhangwaná ni-ré-yé mlóndo Nkangwana COP-BE-REL:l Mlondo 'A Nkangwana is indeed a Londo person.'

(NMb)

(5.128) ve-re-hó úko ve-si-húndúk-e 2-BE-REL:16 there 2-NEG-return-NEG 'Some didn't return from the war, ...'

vitá war(Sw)

(NMb)

The verb óka (NMb) has an equivalent kwá in IMb, which is realised as xwá in the Bumbuli dialect. The verb is used with qualitative complements such as adjectives ((129), (130)) or nouns or noun phrases expressing a quality ((131), (132), (133)). The complement can also be a quantifier ((134), (135)). (5.129) vinyongo vy-óoka vi-dorí insects 8-be 8-small 'Insects are small.'

(NMb)

(5.130) n-áa-kwá mw-ezá 1SG-PST-be 1-tall 'I was the tallest.'

(IMb)

THE VERB

151

(5.131) umeme w-a-oká határi lightning 14.2-INC-be danger 'Lightning is dangerous.'

(NMb)

(5.132) w-ók-e mw-angalizi 2SG-be-SBJ 1-careful 'Be careful!'

(NMb)

(5.133)

(IMb)

idí l-áa-kwá lá vaMa'á land 5-PST-be 5:C0N Mbugu 'The land was that of the Mbugu.'

(5.134) mbú j-óoka ny-ingi mosquito 10-be 10-many 'Mosquitoes are many.'

(NMb)

(5.135) vw-áa-kwá vu-gheré 14.2-PST-be 14.2-few 'They were few.'

(IMb)

A range of other complements of these verbs 'to be' is possible, e.g., indicating man­ ner ((136), (137)). The verb can also be used with inchoative meaning 'to become' ((138), (139)). (5.136) ni-hó ni-'élezea ka-háli chó y-é-kwá-e. COP-16-REL 1SG-explain time-other 7:-REL 9-PST-BE-PF 'I explained again how it was.'

(IMb)

(5.137) íngeí ulojó ú-ne-oka ingeí now harvest 3-FUT-be now 'There was really a harvest now.'

(NMb)

kabísa completely(Sw)

(5.138) basi v-áa-zóko, v-áa-kwá vényéji. well 2-PST-stay 2-PST-be locals 'They stayed there, they became familiar.'

(IMb)

(5.139) ku-óka m-dóri 15-be 1-small 'To be(come) small.'

(NMb)

The verb can be followed by SC-na ((114)), or simply by na 'with', e.g., va- na mkirindi 'They had a king', indicating association.

152

CHAPTER 5

(5.140) vé-ku-óka vé-na 2-COND-be 2-with cattle 'Them being with cattle,'

ŋombe.

(NMb)

These verbs "to be" are used in combination with conjugated verbs to form compound "tenses", i.e. there are two verbs available for the "tense" prefixes. The verb "to be" is the first verb in these compound tenses. For example, a combination of the verb "to be" — either without any "tense" marker or with the past marker — followed by a verb form with a conditional prefix ku expresses Past Habitual( ( 1 4 1 ) ,(142)). (5.141) mw-óoka m-ku-téma miti 2PL-be 2PL-COND-cut trees 'You used to cut trees.'

(NMb)

(5.142) áa-óka a-ku-s(ú)wa ku-nwá mawå l:PST-be 1-COND-refuse 16-drink beer 'He used to refrain from drinking.'

(NMb)

Locative verbs áho (NMb) and åta (IMb)'. The Normal Mbugu locative verb áho is in form identical to the locative class 16 second degree demonstrative, see 5.4, which can also be used adverbially with the meaning "here". In the present work I gloss it as HERE. The complement of áho and ata can be "here" again, i.e. i,i and áta in NMb and IMb, respectively (143), (147). The data on its verbal usage are limited; all available sentences are presented below. It is not clear to which extent it can take "tense" pre­ fixes. In all examples (143)-(147), the complement is locative. (5.143) é-re-áta Vi 1-BGND-HERE here 'He was here.'

(IMb)

(5.144) e-(a)ho kuúme k-ángu 1-HERE right 15-my 'He is on my right.'

(NMb)

(5.145) basi ni-hé hanhu ve-ó-ahó vá-ra-vuka aha aha, (NMb) well(Sw) is-16 place 2-REL-HERE 2-BGND-leave here here 'Well, they left the place where they were, ' (5.146) vinyama vyá ngunjini vi-(a)ho nihío animals 8:CON forest 8-here where 'Where are the animals of the forest?'

(NMb)

THE VERB (5.147) ve-áta 2-HERE

áta

153 (IMb)

here

'They are here In one example ((148)) the complement of the Inner Mbugu verb is locative in form but not in meaning and when it occurs without a complement, the verb indicates pres­ ence ((149)). Furthermore, the Inner Mbugu verb áta is also used with non-locative complements such as a numeral complement ((150)) and a pre-posed manner com­ plement ((151)), and with the preposition na to express possession ((152)). Some of these usages of the Inner Mbugu verb áta overlap with the functions of the kwa verb 'to be' and the SC-na construction, while the Normal Mbugu verb aho seems to be re­ stricted to locative functions. (5.148) i'i hódi ni-áta lusótó mira i'i ni-áta he-nyamálo ghó. here place 1SG-HERE Lushoto but here 1SG-HERE 16-work my 'I am in Lushoto; but I am here for work.' (IMb) (5.149) maganga á-he-líta s-é-áta maganga 3SG-PST.COND-come 1SG:NEG-PST-HERE 'When Maganga came I was not there (present).'

(IMb)

(5.150) kái mu-áta vahé if 2PL-HERE people 'If you are ten people,'

(1Mb)

(5.151) ká-cho

ikádo ten

chi-re-áta

(IMb)

DEM2-7:REL 7-BGND-HERE

'It is how it is,' (5.152) mu-áta na afá 2PL-HERE with goats 'If you have ten goats,'

ikádo ten

(IMb)

Verbal derivation Verbal derivation mainly affects verbs by changing their meaning and as a cor­ ollary, their argument structure. There is no derivation from a noun to a verb; only two examples of verb stems that are derived from noun stems can be mentioned: makáhlúwa 'to dream' [IMb], derived by a passive from the noun makahlú 'dream' (IMb)17, and hlómi 'to anoint' [IMb], derived from ma-hloma 'fat, oil' [IMb], as in (153). 5.6

154

CHAPTER 5

(5.153) m-borá é-tá-kú-hlómi mahlóma girl 3SG-EVI-REFL-anoint fat 'The girl anoints herself with oil.'

(IMb)

Mbugu verbs can be derived through the following verbal derivational suffixes or ex­ tensions, all of which are common Bantu derivational suffixes with their usual func­ tion: suffix -ish -ij -i -ik -an -any -w Table

gloss INT CS APL ST RC RC+APL PAS

name intensifier, causative causative applied stative reciprocal

passive 5.13: Verbal derivational suffixes

What can be stated about the order of these derivational suffixes is that the passive is always the last in the row, as in ni-kúrú-í-w-a 1 SG-cultivate-APL-PAS-F 'I am culti­ vated for'. Some combinations of extensions have become fixed, e.g., -any RC:APL, and -ikan ST:RC. The suffixes are added to the stem, i.e. the verb form without the in­ flectional final vowel. Thus the inflectional final vowel is moved to a position follow­ ing the derivational suffixes.18 After the addition of a derivational suffix all stems end in an inflectional final vowel, i.e. also those which have a root ending in a noninflectional final vowel, as for example in ku-kúrú-í-w-a 15-cultivate-APL-PAS 'to be cultivated for', from ku-kúru 'to cultivate' (IMb). The perfect suffix -iye is reduced to -e after extensions. The tone on the vowel of the extensions is identical to the last tone on the stem. When a suffix is added to a verb root that ends in a vowel (i.e. a non-inflectional vowel), the result is a sequence of which the second — that is, the vowel of the verbal suffix— is sometimes elided, e.g., mátú-(í)sh-a 'hide a lot'. The suffixes containing i have lexically conditioned allomorphs with e after some, but not all, roots containing e or o, e.g., ong-ej-a 'cause to add' but om-ij-a 'cause to dry', and kweny-er-a 'be bitter' but kel-il-a 'to pass sth'. All verbal derivations show up in some lexemes with a lexicalised meaning, e.g., li-sóísh-e 5-go:CS-PF 'it measures', literally 'it causes to go to'. The frequency of certain other verbal endings is so low in my material that they can hardly be treated as suffixes; in shape they seem to contain an extension but one for which there is no pro­ ductive usage. A number of these endings are known from other Bantu languages, e.g., uk in húnd=úk-a 'return', which also appears in loan verbs containing the Bantu ex­ tension, e.g., kúmb=úk-a 'remember' from Swahili. Historically of potentially crucial

THE VERB

155

interest are the non-productive verbal elements such as ti, ri and others in Inner Mbugu that do not resemble the common Bantu extensions. Because of their historical significance these non-productive verbal elements are discussed in detail in section 3.6. The intensive/causative -ish: The suffix -ish has an intensifying meaning and is ex­ tremely productive; the suffix also functions as a causative, from which it developed, e.g., hlétísha 'be very late, cause somebody to be very late'. If the final consonant of the verb stem is k, the suffix manifests itself by a change from this  to sh. This is not to be interpreted as a process which transfers kish to sh, but rather as a reinterpretation of an allomorph of a former causative suffix -i which evoked the palatalisation process k-i → sh; the same process that gave rise to the suffix -ish. Historically the change from  to sh does not come from a suffix ish but as a result of the identity in form; the meaning of intensity is linked to this morphological process of k to sh and therefore in these cases the change k → sh should be considered as an 'allomorph of the intensive suffix -ish. base meaning derived form meaning language juta pull jutisha I+NMb pull hard bigha hit bighisha hit hard NMb im stand imusha stop NMb kúru cultivate kúrúísha cultivate a lot IMb hléti be late hlétísha be very late IMb wear 'óro 'óróísha wear a lot IMb mátu mátú(í)sha hide hide a lot IMb dáha walk dáhásha walk a lot IMb Table 5.14: The forms of the intensive/causative -ish All these intensive verbs end in sha, the consonant of the suffix plus the inflectional final vowel. As a further development, there is a morpheme of the same shape, sha, with an intensifying meaning that cliticises to words other than verbs, such as adjec­ tives, e.g., úda (adj) 'far' údasha 'very far' (IMb), bosísha 'the very first' from bosí 'first' (IMb). The causative -ij: The other productive causative suffix — but with no intensifying meaning — is -ij. Again, the vowel of the suffix is sometimes dropped after noninflectional root vowels, as in kúru, kwáha and 'óro in Table 5.15, but also in the case of ika-a.19 The effect of the suffix on stem final n is to make the nasal palatal, as in vinya and zanya in Table 5.15. Historically, the palatalisation of the nasal arose from a causative suffix i, that synchronically only survives in this change and in the fact that kuhunija can exist next to hinya, both ultimately from the root hina.

156

CHAPTER 5

base hina ikaa ró kúru kwáha 'óro vina zana hina

meaning derived form meaning kuhinija20 learn learn (oneself) sit ikaja put to sitting wait róíja refuse sb let sb cultivate cultivate kúrú(í)ja kwáhája make tired tired wear 'órója put on hurt vinya get hurt meet zanya collect learn hinya teach Table 5.15: The forms of the causative -ij

language I+NMb NMb IMb IMb IMb IMb NMb NMb NMb

The causative is semantically empty in verbs such as fi(r)íja 'to finish' from fí 'to fin­ ish' (IMb). Causatives can follow a non-productive causative as for example in kuhlétíja 'to cause to be late' from kuhléti 'to be late' (IMb), and kuhlátíja 'to cause to close' from kuhláti 'to close' (IMb). A double causative occurs in se-j-ij-a 'pregnant (of cow)' from se-j-a 'put down' from se-a 'descend' (NMb); the IMb equivalent contains an applied suffix after a non-productive causative. Applied extension -i: The applied suffix is -i; it is y for verb stems that end in a noninflectional vowel, as in hlamu, zéu, dára and gwá below. If this final root vowel is an i, as in hláti below, the applied is recognisable in the fact that the verb now has an in­ flectional final vowel a; the verb ends in ia just like regular verbs with a consonant final root. In the case of HL verb stems, the final vowel of the stem now becomes High as well, as if it were the vowel of the extension. Examples of the applied exten­ sion are given in Table 5.16; sb = somebody. base puta luka hlamu zé dára gwa  hláti

meaning derived form smoke putia twist, plait lukia anoint hlamuya herd zéúya grasp dáráya steel gwáya beat maya close hlátía Table 5.16: Applied

meaning smoke for sb. twist for sb anoint sb herd for sb grasp for sb rob sb peg, beat at close for sb extension

language NMb NMb IMb IMb IMb IMb IMb IMb

A lexical variant of the applied suffix is -ir, with its allomorph -il after /, which occurs as a frozen form in a number of verbs such as dáharira 'mould' (IMb), húngíra 'fan, wave' (IMb), kondera 'twist' (IMb), kwenyera 'be bitter' (NMb), towera 'swim'

157

THE VERB

(I+NMb) (loan from Shambaa), izihirwa 'be happy' (NMb), tundirwa 'be taken care of. Some verbs allow either of the applied suffixes with no difference in meaning, e.g., kel-il-a or kel-i-a 'pass (tr)' (NMb) and ger-ir-a or ger-i-a 'bring sb' (NMb). A double applied suffix -ir is contained in jutirira 'pull and pull' from jutu 'pull' (IMb). Other verbs which take the suffix -ir are: fikira 'arrive somewhere' from fika 'arrive,' shíghíra 'leave behind' from shígha 'leave behind', fóríra 'dig at' Worn fóra 'dig'(all NMb). The causative follows the applied in a number of derived stems such as shig=ir-ij-a 'cause to leave behind'. The applied has the form -ir in this combination which makes the two suffixes easily recognisable. The verb 'óró-í-j-a 'to dress sb' (IMb) contains a combination of an applied and a causative suffix, cf. 'óró-j-a 'to put on clothes'. The meaning of the applied extension is that the action is aimed at the object ((154)), which is most often beneficiary ((155)) or a location ((156), (157)). The ap­ plied creates an object which is usually different from the object of the base verb; thus it renders intransitive verbs transitive. The object can consist of an object prefix only, ((154), (155)). (5.154) é-ta-m-gwá-y-a wá 1-EVI-1-steal-APL-F cattle 'He steals cattle from him.' (5.155) ku-ní-gér-i-a ka 15-1SG-bring-APL-F time tobacco 'to bring me some tobacco'

(IMb)

wa'ú

(IMb)

(5.156) ni-kel-il-é hépara 1 SG-pass-APL-PF KwePara 'I passed Kwepara.'

(NMb)

(5.157) hé-fórírwe ishima 16-dig-APL-PAS-PF pit 'There had been dug a pit.'

(NMb)

The stative -ik: The form of the stative derivational suffix is -ik. The vowel of the suf­ fix is optionally deleted after a non-inflectional root-final vowel in Inner Mbugu. base vóka manya

meaning start know

derived form vókóka manyika

meaning be started be known

language NMb NMb

158

CHAPTER 5

base xá'a kúru 'óro vota

meaning tear cultivate wear win

derived form xááíka kúrú(í)ka 'óró (i) ka voteka Table 5.17: Stative

meaning be torn be cultivated be wearing admit guild extension

language IMb IMb IMb NMb

The stative is used to express the state or possible state of the action or event of a verb without implying the involvement of an agent, cf. 'seen' or 'considered' in (158), and 'audible' in (159). (5.158) kwátómpaká wa'íijí vamaá te-vé-áhó-ik-a ní vahé thus until now Mbugu NEG-2-see-ST-F COP people ve-ló maána 2-have meaning 'Thus until today the Mbugu are not seen as serious people.'

(IMb)

(5.159) lukalasú lú-ko-ik-e kwa cries 11-hear-ST-PF with 'The cries are audible from afar.'

(IMb)

úda far

(5.160) luhíge lw-áa-hlá-ík-a door 11-PST-open-ST-F 'The door is open(ed).' (Sw: mlango umefunguka)

(IMb)

(5.161) mwiko w-a-on(i)-ik-a. taboo 3-PST-spoill-ST-F 'The taboo is broken.'

(IMb)

The reciprocal is expressed by the derivational suffix -an. base juta té kwáha

meaning pull cut tired

derived form meaning jutana pull against each other téána fight each other kwahaána contend Table 5.18: The reciprocal

language NMb IMb IMb

The reciprocal indiates that the action involves various agents and patients interacting with each other, as can be seen in (162), (163) and (164). There is no further object allowed. However, the subject can be singular if the other agent/patient is expressed in a prepostional phrase with na 'with, and, by', as in (165).

159

THE VERB (5.162) v-áa-té-án-a kwa mhpambá 2-PST-beat-REC-F with machete 'They fought each other with the machete.'

(IMb)

(5.163) kaé té-m-ló ku-ró-án-a. longtime NEG-2PL-have 15-leave-REC-F 'You are not to leave each other for a long time.'

(IMb)

(5.164) vahé vé-huru-án-a hé-lumuhlú v-áa-kwáha people 2-pull-REC-F 16-rope 2-PST-tired 'The people who were pulling the rope are tired.'

(IMb)

(5.165) y-á-jut-an-a na ú 9-PST-pull-REC-F with 1 'It pulled him.'

(IMb)

The combined reciprocal-applied suffix -any: This is a fixed combination because it is present in a number of verbs where it has a meaning that cannot be considered as the sum of the meanings of its components. base ghava káka vwija kela fúnya li runga

meaning divide divide ask pass take out come from let sb in

derived form meaning ghavanya divide kákáánya divide vwijanya ask each other sth kelanya pass swh fúnyánya come forward swh liijanya come out rúnganya mix koranya exchange vunganya collect ringanya (nhkondo) prepare (war) Table 5.19: The extension -any

language NMb IMb NMb NMb NMb IMb NMb IMb NMb NMb

The meaning of this combination of extensions is supposedly reciprocal with an ob­ ject, as in (166), or at least a patient, as in (167). A similar basic meaning can also be assumed to be present in the frozen forms koranya 'exchange' (IMb), vunganya 'col­ lect' and in rúnganya 'mix' as in runganya mahemba na maharage 'mix maize and beans'. In the other examples with the combination -any, however, there is no recip­ ­rocity, ((168), (169), (170)), and the combination seems to function as an applied ex­ tension.

160

CHAPTER 5

(5.166) tu-kí-za ha-la tu-vwíj-any-a msí 1PL-COND-arrive 16-DEM3 1PL-ask-REC:APL-F 'When we arrive we'll ask each other about the day.'

(NMb) day

(5.167) á-la va-tátu ve-ghav-any-a 2-DEM3 2-three 2-divide 'Those three divide themselves.'

(NMb)

ve-nyéve 2-self

(5.168) tu-kél-ány-a ndia lPL-pass-REC:APL-F way 'We passed the way.'

(NMb)

(5.169) é-ku-tónga é-ku-funy-any-a áha 1-COND-go l-COND-come:forward-REC:APL-F here 'When he went on and came forward, '

(NMb)

(5.170) shuhé sí-hó i-lí-ij-any-a leopard NEG:COP-16-REL 9-come-CS-REC:APL-F 'The leopard, didn't show up'

(IMb)

The passive -w: The passive is -w and is suffixed to the verb stem. No other deriva­ tional suffix can follow the passive. base kóma tanga jughua gé gwá iva hé

fika sé ké bá gá'a ohlu hlomi

meaning kill call open open steal steal arrive arrive call put tell kill insult anoint

passive kómwa tángwa jughuwa gewa gwáwa iviwa hewa

meaning be killed be called be opened be opened, allowed to marry be stolen, circumcised be stolen, circumcised slim slim fikwa séwe be called kéwe be put báwe be told gá'áwe be killed ohluwe be insulted hlomiwa be anointed Table 5.20: The passive extension

language NMb NMb NMb IMb IMb NMb IMb NMb IMb IMb IMb IMb IMb IMb

The combination applied-passive is very common because the applied introduces the object that becomes the subject of the passive verb. The last verb in Table 5.21 below,

161

THE VERB waja, is derived from the verb wáa 'be ill'. base si bigha hlá waja

meaning rain hit ache ache Table

derived form meaning sííwa be wet bighiwa be beaten, wet fall ill hlaiwa fall ill waj(i)wa 5.21: Applied-Passive extension

language IMb NMb IMb NMb

The meaning of the passive derivation is to change the verb so that its patient becomes the subject. Very often the verb form is in the perfect, i.e. it has the inflectional ending e instead of the basic a. The agent can be expressed in a prepositional phrase with the preposition ní ((173)). (5.171)j-áa-má-w-e 10-PST-beat-PAS-PF 'they (the cows) are beaten.'

(IMb)

(5.172) yá-wáhá-w-e 6-drink-PAS-PF 'it (water) is drunk'

(IMb)

(5.173) u-ku-dárá-w-e ní 2SG-COND-catch-PAS-PF by flu 'You are caught by a cold.'

luáhe

(IMb)

Non-productive extensions A number of verbs seem to contain a suffix which is no longer productive. Such ele­ ments can be recognised in the structure of the verb stem, e.g., CVCVC, and if they re-occur as the final element. Candidates are those where the element can be shown to be a unit because there are related forms without that element, or those to which a meaning can be assigned. Some of these are usual Bantu suffixes such as the separa­ tive -ul and the stative -uk; these two are so rare that they cannot be considered pro­ ductive. The separative -ul: The suffix -ul can be seen with separative meaning in the verb ghúbúla 'to uncover': compare the same verb with stative suffix ghúb-ík-a 'close, cover' (NMb). It can also be recognised as  in chungua 'to loosen' derived from chunga 'to tie, fasten'. A number of verbs, some of which can be clearly associated with a separative meaning, contain an element ul or ol after  VC root which does not exist independently, e.g., bohola 'scoop out' (IMb), hárúra 'shave, cut hair'

162

CHAPTER 5

(NMb), kómbóla 'scrape out' (IMb,), dahula 'pluck' (I+NMb), ghobola '-skin, strip off bark' (NMb), nyukula 'clean by rubbing' (IMb, NMb), bidula 'turn over (e.g., of soil)' (IMb), shélúla 'ruminate' (IMb,),-togola 'thank, praise' (I+NMb), lojola 'get plenty of food' (I+NMb), gongoloka 'go out of sight over the hill' (NMb). The extension uk: Similarly, a stative extension uk or ok is present in imüka 'stand' (NMb), gongoloka 'go out of sight over the hill' (NMb), húndúka 'return' (NMb), junguluka 'go round', kúmbúka 'remember' (NMb), ódóka 'cold, wet' (IMb), rongoka 'straight' (NMb), sheruka 'boil (intr)' (NMb), umbuka 'have leprosy' (NMb,IMb), and in jugh-uk-a 'be open'; cf. jughua 'open'(NMb). There is a remarkably high number of (petrified) causative extensions in Inner Mbugu that are absent in Normal Mbugu in addition to the productive causative -ij. The one that comes nearest to being a productive suffix is -ti, but there are other candidates such as -ri, -li, and -si, which are less common but which still can be associated with a causative function. All these have the effect of moving the high tone to the penulti­ mate syllable, with a few exceptional cases where high tone is moved to the final syl­ lable. In addition, there seem to be causative elements of the form kV, tV, T a n d rV in which V is a copy of the preceding vowel. Because of the crucially important histori­ cal nature of evidence of non-Bantu morphology; these endings are discussed more extensively in section 3.6.

6

6.1

THE NOUN

The noun class system Both Inner and Normal Mbugu have a Bantu noun class system. Every noun falls into one of the 16 classes which are defined according to agreement phenomena in, among others, adjectives, connectives, and subject and object marking in the verb, see Table 6.1. The classes are to a large extent overt in form because the nouns contain noun class prefixes. Classes 1 and 3 have identical noun class prefixes, but differ in agreement. Class 1 contains only human beings. Otherwise noun class membership does not correlate in any predictive way with meaning. However, derivational place­ ment into a noun class is meaningful, see section 6.4. The classes form regular singu­ lar-plural pairs, as can be seen in Table 6.2. The conventional Bantu noun class num­ bering is used. Classes 14.1 and 14.2 both derive from class 14 historically. Several noun classes contain prefixless nouns; these nouns are considered as a special group within the same noun class and they are referred to by adding a to the noun class number. In the sections on the individual noun class pairings, all these cases will be discussed. The noun class system is identical to that of Pare with the exception of Mbugu class 14.2 which does not occur in Pare and is due to Swahili influence. Lexical items in Mbugu class 14.2 that have equivalents in Pare occur in Pare in class 14, which in turn is the equivalent of Mbugu class 14.1. The nouns in Mbugu class 14.2 have a vowel initial stem. In other words, Pare vw corresponds to Mbugu uw: uwáji (14.2) 'disease' (NMb) from Pare vwáji (14), and uwongó (14.2) 'brain' (NMb) from Pare vuongo (14). Equivalent nouns are members of the same class in both varieties, Inner and Normal Mbugu, and correlate roughly with the lexical contents of Pare noun classes. The class system differs substantially from that of Shambaa: Shambaa has wa for class 2, no prefix for class 5, no class 11, no class 14.1, no class 14.2, and two ad­ ditional locative classes.

164

CHAPTER 6

Pronominal prefix Verbal subject prefix Noun prefix ú é m va vé va m ú ú í 4 mi í 5 i li li 6 ma é à chí 7 ki chí 8 vi ví vi í ì 9 (N) 10 (N) ji jí 11 lu lú lú 12 ké ka ká 14.1 vu vú vú 14.2 ú ú  kú 15 ku kú (ku) hé há 16 Table 6.1: Mbugu noun classes (N refers to a nasal element) Class 1 2

6.2

The morphophonology of the noun class prefixes Table 6.1 gives the basic forms of the noun class prefixes and the agreement prefixes. The following regular allomorphs occur (further details of allomorphy are given in the relevant paragraphs): The high vowels of the noun class prefixes (classes 1, 3, 5, 11, 14.1, 14.2, and 15) become a glide when followed by a vowel, e.g., mw-ahla (3) fire', lw-atame (11) 'light', y-asemi (5) 'breast (female)'. The prefix m is mu before a vowel, h, nasal, or w: muhási 'bed' (IMb), muhati 'young sheep' (NMb), muhibé 'arrow' (IMb), m(u)hatú 'tree' (IMb), munyu 'salt' (NMb), muwa(')ú 'smoke' (IMb). It is m or mu before some fricative initial and r-initial stems: m(u)ghongo 'back' (NMb), m(u)si 'pestle' (I+NMb), m(u)ru'ú 'sweat' (IMb), m(u)ruke 'sweat' (NMb). 6.3

Noun class pairings Noun classes form singular/plural pairs. The pairs in Table 6.2 are common to both Inner and Normal Mbugu. The same pairings occur in Pare. The sequence 6+ in­ dicates that the class 6 plural prefix is added to the singular noun in front of its noun prefix. The locative class 16 has no plural equivalent. Some other pairs occur. Several nouns occur only in a singular or plural class. This is particularly true for class 6 with liquid substances, for classes 11, 14.1 and 14.2 with abstract notions, and for class 15 with infinitives. Details are mentioned in the paragraphs on the various class pairs. Note that class 14.1 occurs both as a singular and as a plural class.

165

THE NOUN

singular plural 1 2 3 4 5 6 7 8 9 10 11 10 11 6+11 12 14.1 14.1 6 14.2 6+14.2 15 14.1 15 6 Table 6.2: Singular/plural pairs of noun classes Class 1 m-; plural: Class 2 va-: This class pair contains human beings only. Normal Mbugu and Inner Mbugu members are parallel. Some examples are given in (6.1). IMb

(6.1) meaning person woman girl child Mbugu person

NMb

SG

PL

SG

PL

m-hé m-lagé m- 'aleta m- 'inyí - 'á

vahé valagé va'aleta va'inyí va-Ma 'á

m-nh(t)u m-cheku m-borá mw-ana m-bughu

va-nh(t)u va-cheku vaborá va-ana va-mbughu

Class la 0; plural: Class 2 va-: Some of the words that have the same agreement as class 1 occur without a noun prefix. These words denote family relationships, e.g., lagé 'mother' (IMb), and are considered to be a semantically and morphologically de­ fined subclass, 1a, with a zero allomorph as the noun prefix. Their plural is not differ­ ent from other class 1 words. Most words in class 1a are common to both languages, IMb and NMb. The complete list of members of this subclass is given (6.2) below. In IMb the word for king, diláo, is in la, although it is obviously not a family relation, while its NMb counterpart, m-fumwa, is in 1 and not la. The words for 'mother' and 'elder brother' show variation between the presence and absence of the prefix, i.e. be­ tween class 1 or 1 a. (6.2)

Singular/plural class pair la/2 containing family relations abá / va-abá 'father' vava / va-vava 'father' baba / va-babá 'father' igíi / va-igíi 'his father, brother'

(IMb) (NMb) (I+NMb) (IMb)

166

CHAPTER 6 vavée lagé I va-lagé (m)-lala ihlíi, ahlíi [mée] abú / va-abú énga / va-énga káká / va-káká kokó / va-kokó ayá kokóye kokorúu / va-kokorúu kákárúu / va-kákárúu wáu (m)-kulúo / va-kulúo egú / va-egú iŋá, / va-iŋá

iŋi(nh)tá / va-iŋi(nh)tá iŋinhtíi / va-iŋinhtíi rumbúria / va-rumbúria rumbúrio / va-rumbúrio

'his father' 'mother' 'mother 'his/her mother' 'his/her mother' 'maternal uncle' 'paternal aunt' 'grandfather' 'grandmother' 'grandmother' 'granddaughter' 'granddaughter' 'grandson' 'sister' 'elder brother' 'his brother' 'my sibling of opposite sex' 'my sibling of opposite sex' 'his/her sibling of opposite sex' 'my sibling of opposite sex' 'his/her sibling of opposite sex'

(NMb) (IMb) (NMb) (IMb) (NMb) (I+NMb) (I+NMb) (I+NMb) (NMb) (IMb) (NMb) (IMb) (IMb) (I+NMb) (NMb) (IMb) (IMb) (IMb) (IMb) (NMb) (NMb)

The cognitive model of the family relations is clearly derived from Bantu: the NMb terms correspond in form and meaning to Pare and the IMb terms are semantically parallel to NMb, showing the same semantic distinctions and groupings. The posses­ sive endings in Inner Mbugu on some of these family relationship terms are of poten­ tial historical interest, see 3.5. They are parallel to those in the Normal Mbugu equiva­ lents but different in form: compare,for example, the ending íi for third person singu­ lar possessive and the ending å for the first person possessive as they appear in iŋinhtá 'his/her sibling of opposite sex' vs irjinhtíi 'my sibling of opposite sex'; and íi occurs also in igíi 'his father, brother' and ihlíi 'his/her mother'. These lexicalised endings do not occur elsewhere in the language. Class 3 m-; plural: Class 4 mi: Class 3 has the same noun prefix as class 1 but differs in verbal concord, in plural pairing, and in semantic content: the lexical items in class 3 are not human beings. (6.3) meaning head tree

IMb SG/PL m-'á / mi-'á m(u)-hatú / mi-hatú

NMb SG/PL m-twi / mi-twi m-ti / mi-ti

A restricted number of nouns in class 3 have no prefix. This is partícula' ly the case of

THE NOUN

167

Inner Mbugu in the Bumbuli area. There can be no doubt about the class membership of these nouns since from their agreement it is evident that they belong to class 3. Moreover, the plurals have the usual class 4 prefix, for example áhla 'fire' (IMbBumbuli), mu-áhla (IMb-Magamba), plural mi-ahla. The same phenomenon occurs with class 14.1 and is considered by Meinhof (1906:302-303) and others after him to be a sign of the instability of the noun class system in Inner Mbugu. This phenomenon and its historical relevance are discussed in full in section 3.5. Class 5 i-; plural: Class 6 ma-: Class 6 contains a number of words denoting liquid substances. It is the plural class par excellence: it is not only the plural class for class 5, but also for class 14.1 and it is one of the plural classes to class 15; the prefix of class 6 is also used as an additional prefix for the plural of certain words in classes 11 and 14.2. Some examples of nouns belonging to the class pair 5/6 are: i-gondwa / magondwa 'lizard' (I+NMb), i-tunda / ma-tunda 'fruit' (I+NMb), i-'ilá / ma-'ilá 'eye' (IMb), i-dahó / ma-dahó 'window' (IMb), ma-wa 'beer' (NMb), ma-la 'beer' (IMb). The class prefix is izi- rather than i- for the stem iso 'eye' in NMb iziso, p1: meso (cf. S. Pare iziso, pl: mazíso). This is a relic of an allomorph before (near-) monosyllabic stems; in other words in class 5 that start in izi, the zi is part of the stem, as for example in izina, p1: mazina 'name' (NMb). Class 7 ki-, plural: Class 8 vi-: The members of class 7 and 8 are parallel in Inner and Normal Mbugu and identical to Pare. The form of the prefix is ki and vi even before vowel-initial roots, with the exception of chumba (7/8) 'room' (I+NMb), which could be a loan from Swahili; other nouns that start in ch belong to class 9/10. Some examples of 7/8 nouns are: kihlatu / vihlatú (IMb) = kichaa / vichaa (NMb) 'finger'; kigi / vigi (IMb) = kinhtu / vinhtu (NMb) 'thing'; kikíre / vikíre (IMb) = kichumbi / vichumbi (NMb) 'chair'; kiete / viete (IMb) = kihojá / vihojá (NMb) 'market'; and kiumbi / viumbi 'hunchback' (IMb). Class 9 nasal, plural: Class 10: nasal: Singular and plural noun forms are identical for this class pair. The distinction between class 9 and its plural, class 10, is clear from agreement phenomena. The nasal element of the class 9 and 10 prefix consists of prenasalisation together with stop formation of stem-initial obstruents. The nasalisation is voiced with voiced stops and voiceless with voiceless stops. The voiceless prenasalisation renders the following voiceless stop strongly aspirated, or, alternatively, the stop is not realised at all (see section 4.6). Since the noun prefix is identical in sin­ gular and plural in this pair, it is impossible to determine the stem without the prefix in most cases. Evidence for the shape of the noun prefix comes from the pairs 11/10 and is presented in section 4.6 and from some rare cases of derivation such as ndoji 'mar­ riage' from dori (v) 'to marry' (see also section 6.4). For words in class 9/10 that start in a plain nasal, it is arbitrary to analyse the prefix as zero or as being contained in that nasal.

168

CHAPTER 6

singular mbuvá nhtembo mhpirimhpiri mbuji nguvu ndeghe nhkingo nhtao nhxwahle manga mindá nyumba nyoka

meaning field elephant pepper goat power, force ndeghe bird nhkingo hide nhtao bull nhxwahle partridge manga cassava míndá house nyumba house nyoka snake Table 6.3: Class pair 9/10 plural mbuvá nhtembo mhpirimhpiri mbuji

language I+NMb I+NMb I+NMb NMb NMb NMb NMb NMb IMb I+NMb IMb NMb NMb

The realisation of the prefix is zero with stems that begin in a voiceless fricative such as s, sh or/, or the IMb fricatives such as hl and x: fanus i / fanusi 'lamp' (I+NMb), fole 'fear' (NMb), hlané / hlané 'male goat' (IMb), sawa I sawa 'louse' (I+NMb), shuhe / shuhe 'leopard' (I+NMb), sa'a 'cold, wind' (IMb), xehlú / xehlú 'kind of rat' (IMb), ximéno / ximéno 'bird' (IMb-Magamba variety). The approximant h also occurs stem initially in class 9/10 nouns without a noun prefix, e.g., himéno / himéno 'bird' (IMb), himé 'dew' (NMb) and the Swahili loan héra 'money (I+NMb). Some of these h-initial class 9/10 words are dialectal variants of x in IMb, or Swahili loans in I+NMb. In class 11/10 pairs we will find nasalisation plus h resulting in mhp next to h in words with a zero prefix. Only in IMb are there class 9/10 nouns that occur with a zero allomorph of the class 9 or 10 noun prefix before a glottal stop and with vowel initial roots: 'u / 'u 'power, 'force', afá / afá 'goat', ahlú / ahlú 'hair', éhla 'tradition, folk-lore', ogú / ogú 'castrated goat'. It is unclear whether the absence of the prefix is morphologically or lexically conditioned: maybe these words belong to the next category. There are some words in 9/10 without a prefix but with stem initial consonants that normally require nasalisation. Some of these can be explained as Swahili loan words, and they occur in the common lexicon (I+NMb). These are: taa / taa 'lamp' (NMb) from Swahili, tumbatu 'tobacco' (I+NMb) (an areal word), wali 'cooked rice' (I+NMb) from Swahili. Two additional NMb words in 9/10 that exceptionally lack the initial nasal are káa 'home' (NMb) (identical in Pare Ma), and chololó 'urine' (NMb) derived from cholola 'urinate'. In the IMb lexicon, more such cases of a lexically de­ termined zero-allomorph exist, probably in analogy with the numerous phonologically determined zero-allomorphs of the prefix in IMb. All examples are given in the fol­ lowing table.

169

THE NOUN

singular plural wa wa buhlé buhlé borné borné chalú chalú kwa kwa gomaé gomaé dísha dísha Table 6.4: Class 9a

meaning language cattle IMb hide IMb IMb snake bull IMb path IMb clothes IMb valley IMb 0, plural class 10a Ø

Class 11 lu-; plural: Class 10: Words in class 11 either have a plural in class 10, Table 6.5, or form their plural by the addition of the class 6 ma- prefix to the class 11 prefix, Table 6.6. The class pair 11/10 enables us to see alternations of absence and presence of a class 10 nasal prefix and thus shows us the phonological processes of nasal plus stem-initial consonant integrating into one syllable onset, see Table 4.11 above. In addition, the following allomorphy can be observed: Vowel initial and monosyllabic stems have a prefix nyi- in class 10. Note that not all cases of root-initial h are ren­ dered mhp as it is presented in the phonology section; in fact, in most cases class 10 has a zero allomorph with h-initial stems and with stems that begin in x (which is sometimes realised as /7), as in lu-ximú, p1: himú 'fingernail' (IMb). Other voiceless continuants, such as s also show the zero allomorph for class 10, lu-símó, pl: simó 'story, saying' (I+NMb) and lusirá, pl: sirá (11) 'necklace' (NMb).1 Furthermore, the glottal stop in IMb also conditions a zero allomorph of the class 10 prefix, lu-'ára, pl: 'ara 'baldness' (1Mb). singular luambo lute luge lungo lúmi luharemú luhabi luhási luhazi

plural nyambo nyoté nyige nyingo nyimi haremú habí

meaning hide-peg saliva, curse vein, tendon straw plate tongue horn firewood tendon, vein hási hazí baldness Table 6.5: Class pair 11/10

language I+NMb NMb NMb I+NMb NMb IMb IMb IMb NMb

Some words in class 11 have no plural at all. These are: lwatamé 'light' (IMb), luwingu 'cloud' (I+NMb), lukalasú 'cry', luáhe 'cold', lukopa 'hide (prepared)' (NMb), lupási 'animal scent' (IMb), lutéro 'fragrant plant for cleaning gourd' (I+NMb), lu'adáka 'meat of the tongue as far as the neck (cow)' (IMb), lu'atamé 'summer light­ ning' (IMb), lwéngu 'spleen' (I+NMb), lwávú 'net' (I+NMb), luzíngé 'thread',

170

CHAPTER 6

lushungí 'comb, crest of cock' (I+NMb), lukata 'ladle' (NMb), lushofi (I+NMb) 'sharpened wood'.

singular lu'ánda lúmi lu'iréno lubambara lubamé lubégho lubwá'o lughóhe lugo lugumda luhige luhondo lukóho lukólo lukúmbi lumurá lupúra lusha'ú luvó

plural meaning malu'ánda tongue malúmi tongue sleep malu 'iréno malubambara lightning malubamé sole of foot malubégho diarrhoea malubwá'o mountain sleep malughóhe malugo fence malugunda horn maluhíge door maluhondo trough malukóho cough malukólo clan malukúmbi outer wall, courtyard malumurá drill, needle malupúra cough malusha'ú calabash plant maluvó calabash plant Table 6.6: Class pair 11/6+11

language IMb NMb IMb IMb IMb I+NMb IMb NMb I+NMb NMb IMb I+NMb NMb IMb I+NMb I+NMb IMb IMb NMb

Class 12 ka-; plural: Class 14.1 vu-: The main functional load of this class pair is derivational: Diminutives are formed by transforming the noun to this pair of classes. A small number of lexical items occur exclusively in this pair of classes. Most of these denote small entities such as -'inyí 'infant' (IMb), ka-nyaxe 'baby' (IMb-Rangwi variety), ka-tangína 'infant' (NMb), ka-dendé 'small pot for vegetables' (IMb), kahatá 'peg', ka-suhu 'small pot' (NMb), and several birds such as ka-deghé 'bird' (NMb), ka-sio 'chick' (I+NMb), ka-shózi 'sunbird' (I+NMb). Also bigger birds, such as ka-zewé 'secretary bird' (I+NMb), occur in this class pair. Other lexical items that are not obvious exponents of diminutive meaning are, e.g., ka-bugha 'hare' (IMb), ka­ sú 'thirst' (IMb), kiru 'thirst' (NMb) The word ka-zúva 'dry season, heat' (NMb) can be taken to be derived from izúva 'sun, day' (NMb), just like its parallel in IMb ka-'azé 'dry season, heat (sun­ shine)' from m-'azé 'sun, day' (IMb). The derivation, however, does not seem to in­ volve the diminutive. Class 14.1 vu-: plural class 6 ma-: A number of words in class 14.1 have no singular

THE NOUN

171

equivalent. Those that do have a plural, have it in class 6. Some examples are: vunuhéa / manuhéa 'intestines' (IMb), vutumbo / matumbo 'intestines' (NMb), vubasá / mabasá 'face' (IMb), vushó /mashó 'face' (NMb), vukusó /makusó 'bow' (IMb), vutá /mata 'bow' (NMb), vuáho 'porridge' (IMb), vughai 'porridge' (NMb). Class 14.2 -,pluralclass 6 ma+u-: The Pare reflex of Bantu class 14 is vu; the Swa­ hili reflex of this class prefix is u-. Both prefixes occur in Mbugu; the former is re­ ferred to as class 14.1, the latter as class 14.2. Both may occur with the same stem, as is the case with the words for 'brain' and 'illness': vuongó (14.1), ubongo (14.2) or uwongo (14.2) 'brain' (I+NMb); vuwaji (14.1) or uwaji (14.2), pl: mauwaji (6) 'ill­ ness' (NMb); cf. urukao (14.2), pl: maurukao (6) 'illness' (IMb). The fact that the In­ ner Mbugu word parallels the Swahili influenced NMb variant with the u- prefix shows that Inner Mbugu parallel lexicon formation for this lexeme continued until the period of Swahili influence. Plurals are made by adding the class 6 prefix to the full word including the class 14.2 prefix. Some other words in class 14.2 are: ulongó 'clay' (I+NMb), ulojo 'harvest' (NMb). Class 15 ku-; plural: class 14.1 vu- or class 6 ma-: Class 15 contains all infinitives, e.g., kukúru (IMb), kuima 'to cultivate' (NMb). In addition a limited set of non-verbal nouns occur in this class. These have a plural in 14.1 or 6, or in both. The word kunhtu (15) 'manner' (NMb), is derived from kinhtu (7) 'thing'. The following are all nonverbal nouns in the class pairs 15/14.1 and 15/6: kusáme / vusáme, masamé 'leg' (IMb), kughu /vughu 'leg' (NMb), kumoso /mamoso 'left' (I+NMb), kuúmé / maúmé 'right' (I+NMb).2 Class 16 hé+: Class 16 is the only locative class. It is different from the other classes in that addition of a prefix hé- to a noun including its noun prefix makes it a member of class 16, e.g., hélukándo 'on the wall', cf. lukándo (11) 'wall'; the prefix can also be added to a noun phrase (see example (6.4)). A number of nouns, such as hódi (IMb) hanhtu (NMb) 'place', aŋá (IMb) kaa (NMb) 'at home', trigger the same class 16 agreement when they are used as locations, even without the additional he+ prefix. (6.4)

6.4

tu-ta-dáha na hé-m'aŋá 1PL-EVI-walk with 16-3:village 'We walk to another village.'

m-háli. 3-other

(IMb)

Derivation There are two kinds of nominal derivation. Nominalisations of verbs or adjec­ tives to noun and noun to noun derivation within the noun class system. I first consider this latter kind of derivation. The regular cases involve placement into 12/14.1 for di­ minutives;3 into 14.1 (NMb only) and 14.2 for abstracta; and into 16 for locations, see Table 6.7. But other, more sporadic cases of derivation have been attested. The few cases in which a lexical stem occurs in several noun class pairs, e.g., kinhtu (7/8)

172

CHAPTER 6

'thing' next to mnhtu (1/2) 'person' (NMb), are not considered as derivation. Infini­ tives are verbal nouns and constitute class 15, e.g. ku-ima (15) (NMb) 'to cultivate', ku-kúru (15) (IMb) 'to cultivate'. derived noun kahara ka'inyi kaafá kabúji kawahá kahandwí vunh(t)u ugirú uborá ukirindi ulamé hélukándo itúnda ka'azé

class 12 12 12 12 12 12 14.1 14.2 14.2 14.2 14.2 16 5/6 12

meaning base class meaning hara 9/10 calf small calf 1/2 child small child m 'inyí small goat afa 9/10 goat small goat mbúji 3/4 goat knife mwahá 3/4 machete mhandwí 3/4 machete knife thing kinhtu 7/8 such girú adj big bigness girlhood mborá 1/2 girl 1/2 king chiefdom mkirindi falsehood 3/4 lie mlamé on the wall lukándo wall 11 3/4 fruit mtúnda fruit tree 3/4 sun, day dry season m'azé Table 6.7: Noun class derivation

language IMb IMb IMb NMb IMb NMb NMb IMb NMb I+NMb IMb I+NMb NMb IMb

As to nominalisations, one way of nominalising adjectives or verbs is to transfer the stem into one of the noun classes. derived noun mwagirú vukiva ugiru mfwanó

class 1/2 14.1 14.2 1/2

meaning base word class man, elder agirú adj sorrow kiva adj bigness, proud girú adj witch -fwáno v Table 6.8: Nominalisations

meaning male poor big bewitch

language IMb NMb IMb IMb

A number of Inner Mbugu nouns are derived from verbs by means of the suffix -mé: we-mé (14.2) 'price' from wé 'to buy', ki-yo-mé (7) 'word' from yó 'to talk', ki-kurumé (7) 'agriculture' from kuru 'to cultivate', mkurumé (1/2) 'farmer' from kúru 'to cultivate'. Words with no simplex form but which possibly contain this ending are: Iwatamé (11) 'light', lu-'aramé (11) 'broadness'; maybe even i-sheremi (5) 'slipperiness', and the adjective kuhamé (adj) 'black'. The historical relevance of this Inner Mbugu derivational suffix with limited productivity is discussed in section 3.7 above. A small number of Inner Mbugu nouns seem to be derived from verbs by means of a non-productive suffix. These cases are extremely few. We simply state that they exist and present all our examples: -o: m-hlahla'o (3/4) 'dance' from hlahla'a 'to

THE NOUN

173

dance', -mo: m-zemó (1/2) 'herder' from zéu 'to herd', -ji: n-doji (9/10) 'marriage' from dori 'to marry'. Possible historical sources are discussed in section 3.8. 6.5

Frozen suffixes A number of words in Inner Mbugu end in éno, éto, éna or éta. None of these endings is productive. They can be recognised as a detachabl final part in some lexi­ cal items, for various reasons which are: variation within Inner Mbugu with and with­ out the ending, variation between Inner and Normal Mbugu with and without the end­ ing, attestation in earlier sources with or without the ending, and addition of the end­ ing to the historical source of the Inner Mbugu lexeme. The historical relevance of these endings, which look like Cushitic plural suffixes, is discussed in section 3.7. 6.6

Question nouns Some question words can be shown to be nouns on the basis of their syntactic positions. It is however impossible to establish their class membership since they do not trigger agreement and have no plurals with the exception of kini 'what' (NMb), which is a class 7 lexeme that has its plural as vini in class 8.4 Other question words are modifiers (see 7.1 and 7.5). The question words that are nouns are: níjí 'who' (IMb) and nianí 'who' (NMb), ahoní 'what' and (IMb) kini 'what' (NMb), ami 'when' (IMb) and nihi 'where' (NMb). Note that the stem ani 'which' (NMb) is etymologically part oí níaní 'who' (NMb) and kini 'what' (NMb). 6.7

Names Names are nouns from a syntactic point of view. Personal names show the same agreement as other humans and thus could be analysed as belonging to class la. In personal names a distinction can be made between male and female names. Some names have a male and a female version, the male version beginning in se and the fe­ male version beginning in na, e.g., sevuri, navuri. The same phenomenon is found in Shambaa, Pare and other Bantu languages of the region. Names constitute one the se­ mantic fields of the lexicon where there is no distinction between Inner and Normal Mbugu; all names are common to both languages and seem to follow the structure of the surrounding Bantu languages. Male names are: Amashi, Baro, Bĩ, Chadua, Chamare, Chambua, Chamghanda, Chaphoa, Charo, Dafa, Hiza, Hwai, Idawa, Irioko, Ishika, Kajime, Kajru, Kakai, Kambio, Kandieni, Kapera, Kibai, Kibanga, Kibwali, Kibwalo, Kidallah, Kidenga, Kigoi, Kihedu, Kihio, Kimwemwa, Kinyashi, Kiogwe, Kisaka, Kisapha, Kitia, Kitojo, Komba, Kubane, Lushino, Mabaruku, Machaka, Mahonge, Makadi, Manya, Mayunga, Mboko, Mbura, Mchavo, Mgheni, Mhandi, Mhando, Mhina, Mnavu, Mngasu, Mroki, Mshami, Mwambi, Mweri, Mweta, Naragha, Ndaji, Ndalla, Nyaki, Ojoi, Otoni, Otono, Rere, Salu, Segao, Seikumbi, Sekhondo, Semahungo, Semghanga, Semhito, Semkiwa, Semkuya, Semshumi, Semzighwa, Sendara, Sendia, Sendighe, Sengasu, Senya, Semhpombe, Sevuri, Shishime, Shogholi, Shogholo, Singo, Udingi, Vuri.

174

CHAPTER 6

Female names are: Bari, Haraghi, Hela, Hkoba, Kabuji, Kahindi, Khunde, Kideghe, Kione, Kira, Kiri, Maghoba, Majuma, Makihio, Malukindo, Mangereza, Mchikirwa, Mhpito, Mizighi, Mkengwa, Mkiwa, Mhpito, Mshuiwa, Mshumi, Nacharo, Naghondo, Nagonja, Nahwai, Nakhondo, Nakiore, Namavoo, Nambua, Namburi, Namhando, Namhina, Namkeri, Namshamba, Namshitu, Namsi, Namwai, Namwasi, Namweta, Nangasu, Nanguji, Naphoa, Namhpeho, Nasama, Nashika, Navesa, Navo, Navuri, Ndodo, Nembogha, Nemghanga, Nemshami, Nesala, Neshumo, Nesira, Nghalwa, Nguriro.

7

ADJECTIVES AND OTHER NOMINAL MODIFIERS

Most nominal modifiers follow the head noun and agree with it in noun class. Demon­ stratives and quantifiers may precede the head noun. Modifiers can also occur inde­ pendently, i.e. when the head noun is not expressed but understood. Nominal modifi­ cation without noun class agreement does occur as well, for example, the word kira or kila 'every' (I+NMb), a loan from Swahili, does not show agreement and always pre­ cedes the noun. The higher numerals do not show agreement and there is no agree­ ment on any numeral modifying the noun ka 'time (as unit of counting)', e.g., va-hé va-kái 'three men' (IMb) but va-hé ikádu 'ten men' (IMb), ka tatu 'three times' (NMb) and ka ngi 'many times' (NMb). The modifiers that show noun class agree­ ment can be divided into two groups according to the type of agreement. On the one hand, there are adjectives that have noun class prefixes as agreement markers; on the other hand, there are pronouns — relative clause pronouns, possessives, demonstra­ tives, and quantifiers — that use a different but related set of agreement markers, termed pronominal prefixes. 7.1

Adjectives Adjectives show class agreement with their head noun, both when used attribu­ tively and predicatively. Attributive adjectives follow their head nouns. The agreement prefixes are essentially identical to the noun class prefixes; thus adjectives are closer to nouns than to verbs. In two classes there is a difference between the adjectival and the noun prefix: The class 14.2 adjectival prefix is i rather than u; and the locative class 16 adjectival prefix is ha rather than he. We can define adjectives as words that agree with a noun and that basically use the same prefix. These contrast with other nominal modifiers that use a different set of agreement markers (see section 7.5). Adjectives form a closed set and their number is limited. There are no derived adjectives. Most adjectives have a Normal Mbugu and an Inner Mbugu equivalent. Table 7.1 gives a full list of adjectives (= indicates that Inner Mbugu uses the same adjective; sometimes with an archaic alternative). The Inner Mbugu adjectives are semantically equivalent to the Normal Mbugu ones. The colour 'red' is rendered by a verb in Inner Mbugu. The quantifiers 'which?',

176

CHAPTER 7

meaning bad big black a certain complete, all fat, heavy few, small five four fresh (of milk) fresh, unripe good, nice hard, difficult how many little (of water) long, high, tall long (of grass) male many new old one open, uncovered other, else red round, long self sharp short small smart, deceitful thin, short, narrow three two unripe, bad which?2 white

NMb IMb viví kusá girú bahá jíru (kuhamé)1 sikú = kabúna vurútu kudusú gheré bwésu kói sano hái áne fwáha susu kénya = tana kuhló kúji heli ngáhi me shishímshi = kuhlé (archaic) ezá shwaghayú = bwange agirú ngi =, kumúre shá ku'á, kú'ale, ti'ale ? tama mwé wé ájí hlemé hali mhpundu sholisholi = =, salaghú (archaic) enyé káy = fuhi = dórí gitutú nyányi = sisirí =, kuchachó, o kumumú (archaic) tátu kái iri nu visi 'ihlé hamá jéwa kuga'á Table 7.1: All adjectives

'other', 'many' and 'all' have adjectival agreement prefixes in IMb but their equiva­ lents in NMb have pronominal agreement prefixes and therefore are not considered

ADJECTIVES AND OTHER NOMINAL MODIFIERS

177

adjectives in NMb, see 7.5. Thus for the quantifiers there is a difference in agreement behaviour between IMb and NMb and the lexical parallelism between NMb and IMb is not perfect here because of this difference in type of agreement prefixes. The Inner Mbugu adjective for 'one' is derived from its Normal Mbugu equivalent. The Inner Mbugu word for 'male' is derived from the word for 'big'. Reduplication of the adjectival stem expresses intensity, as in mafundisho makáykáy 'serious teachings' (I+NMb). Adjectives can be used on their own when the head noun is understood. For ex­ ample, in my corpus, the adjective háli 'other' is used in class 9, as in háli ya kuyó 'else to say' with no antecedent mentioned, while on another occasion the same adjec­ tive is in class 7, now presumably referring to kinhtu 'thing' in kiháli chá kuongeja 'else to add'(in both cases translations of the utterance 'I have nothing else to say'). The distinction between a noun and an adjective is defined above on structural grounds: adjectives modify nouns. But we have also seen that the noun need not be expressed and that the adjective can occur independently. When nouns modify nouns they are normally preceded by a genitive pronoun (see section 7.2), and thus they are still clearly different from adjectives. However, occasionaly nouns can modify other nouns without such a genitive pronoun in a compound construction, e.g., m-nhtu mkiva 1-man 1-pauper 'a poor man". In this example we could easily mistake kiva for an adjective. The fact that mkiva can stand alone meaning 'pauper' would not change that initial analysis since adjectives can occur without their head noun being present, as was shown above. The limited class distribution of this supposed adjective could be attributed to its semantics and considered coincidental rather than structural. However, the derivation for the abstract noun 'poverty' in class 14.2 is u-mkiva, which is based on a noun including the noun prefix rather than on a prefixless adjective stem. It is for this reason that even in these exceptional cases of compounding the distinction be­ tween noun and adjective can still be maintained. Adjectives modify and agree in class to the noun they refer to, even if that noun can be left unexpressed; stacking of pre­ fixes does not occur. Nouns can be placed in several noun classes but never in all and when they do, their meaning changes, whereas adjectives can occur with all noun class (agreement) prefixes. The number of adjectives being restricted, many adjectival concepts are ren­ dered by verbs. Some examples are: hlagha 'be broad' (IMb), héyu 'be slim' (IMb), ikandane 'be dirty' (NMb), sakája 'wear out', 'aku 'shine, be white' (IMb), bayu 'be red' (IMb), tuhu 'be sharp' (IMb), edi 'be good' (NMb). Qualities such as cattle col­ ours, on the other hand, are nouns: mbalazi 'white spot on the back'. The numbers one to five are adjectives. Higher numbers do not show agree­ ment, that is, they are invariable. It is noteworthy that many adjectives in Inner Mbugu start with kw. kusa, kuhamé, kudusú, kuhló, kuhlé, kumúre, ku'á, kuchachó, kumumú, kuga'á. Apart from the variation in the word for 'new', ku'á, kú'ále or tí'ále, there is no reason to separate this element ku from the adjective stem. See section 3.5 for a discussion of a possible Cushitic source for this element.

178

CHAPTER 7

Another noteworthy phenomenon as regards the fonn of Inner Mbugu adjec­ tives is the fact that most polysyllabic stems have a tone pattern of a sequence of Low tones and a final High tone, L--LH. This tone pattern is discussed in section 4.5. 7.2

Genetive and relative pronouns Genitive pronouns (also termed connective or associatives) represent a head noun that is modified by a second noun; relative pronouns are modified by a verb (i.e. a relative clause) (see Table 7.2). The noun class agreement morphemes in these pro­ nouns are the pronominal prefixes, which are different from noun class agreement pre­ fixes in adjectives and nouns (see Table 6.1). class Pronominal Prefix genetive pronoun: -á relative pronoun: -ó 1 u wá wó 2 vá vá vÓ 3 ú wá wó 4 i yá yó 5 lá lá ló 6 á á ó chá 7 chí chó 8 ví vyá vyó 9 i yá yó 10 jó já jí 11 lú lwá lwó3 12 ká ká kó 14.1 vú vwá vó 14.2 ú wá wó 15 kú kwá kó 16 há hó há Table 7.2: Genitive and relative pronouns The genitive pronouns consists of the stem á, preceded by the agreeing pronominal prefix, as in (7.1) and (7.2), glossed with "of; the relative pronouns have the same pronominal prefix plus the stem ó (see (7.3) and (7.4)), and are glossed with "that". The stem vowels coalesce with the vowels of the pronominal prefixes. The repre­ sented noun need not be present in the sentence, as in (7.5). These pronouons cliticise to the right, thus, to the modifier. The relative pronoun is also used after the general purpose preposition na: marukáo ó niló na ó ni makái 'the illnesses that I have are se­ rious'. When used with this preposition na, it cliticises to the left, to the preposition. (7.1) ki-gi yá ni chá-shaba 7-thing this is 7:of-copper 'This thing is of copper.'

(IMb)

ADJECTIVES AND OTHER NOMINAL MODIFIERS

179

(7.2) mbúji ya-mgheni yáafwá 9.goat 9:of-1-guest 9:FPST:die 'The guest's goat has died/

(NMb)

(7.3) vi-aghú vyó ni'á ni 8-food 8:that lSG:eat is 'The food that I eat is nice.'

(IMb)

(7.4)

vi-kuhló 8-nice

vi-jú vyó nilá ní vi-tana 8.food 8:that lSG:eat is 8-nice 'The food that I eat is nice.'

(NMb)

(7.5) hó m-tónga ni haé 16:that 2PL-go is far 'Where you are going is far.'

(NMb)

7.3

Possessives The possessives differentiate person and number but not sex (see Table 7.3). In Inner Mbugu, the possessives show no agreement when immediately following the head noun attributively. They do show agreement, however, when used predicatively ((6)), or when used attributively after another non-agreeing modifier such as a demon­ strative ((7)). This agreement behaviour of possessives is markedly different from that in NMb and in Bantu in general. The Normal Mbugu possessives agree with the head noun in all types of constructions ((8), (9)) whereby the head noun need not be pre­ sent. The agreement, for both Inner and Normal Mbugu, is marked by an element which is identical with the pronominal prefix, except for the facts that the tones are all low and the class 6 prefix is ma, not a, ((7)). NMb IMb -angú ghó -akó or -a-ghó ké -akwé kú'u -etú kánu -enyú kághu -avó kíni Table 7.3: Possessive (7.6) ki-sima yá ni ki-kánu 7-well this is 7-our 'This well is ours.'

meaning 1SG 2SG 3SG 1PL 2PL 3PL

stems

vs. ki-sima kánu 7-wel our 'our well'

(IMb)

180

CHAPTER 7

(7.7) ma-ginera yá ma-ké vs. ma-ginera 6-roots this 6-2SG.POS 6-roots 2SG.POS 'These roots of yours.' 'Your roots.' (7.8)

ké

m-zi w-étu úu 3-town 3-our 3:DEMl 'This town of ours.'

(7.9) i-sima ili ní 5-well 5:DEMl is 5-our 'This well is ours.'

(IMb)

(NMb)

l-étu

(NMb)

The Normal Mbugu possessive stems mostly begin in a, which is the genitive pronoun going back to a structure of genitive pronoun followed by a personal pronoun. The Inner Mbugu stems begin in k, which is probably a relic of the Cushitic mascu­ line/neuter marker ku (see section 3.9). The Normal Mbugu possessive stems have a tone pattern LH; the Inner Mbugu stems are H or HL. Third person possessives may refer to non-humans as in hú buhlé yakwé 'this hide of it (=a bull)' [IMb]. In associative constructions in which the second noun is a personal name, the third person possessive is often used instead of the genitive pro­ noun: nafási yakwé térens 'the position of Terence' [NMb], nyumba yakwé nambai 'the house of Nambai' [NMb], hákwe kidalá 16:3POSS . 'Kidala's place' [NMb]. 7.4

Demonstratives The demonstratives have three degrees of deixis. In Inner Mbugu the demon­ stratives are invariable: yá 'this', kå 'that (near you)', hú 'that (distant)'. In Normal Mbugu the demonstratives agree with the head noun by using the pronominal prefix. As was the case with the possessives, the class 6 prefix is ma-. The NMb demonstra­ tives all have two moras: CVCV, VCV or VV. The proximal demonstrative consists of the pronominal prefix preceded by a copy of the vowel of that prefix and a HL tone pattern superimposed on it. The second deixis demonstrative differs from the first one by having an additional final vowel o, with which the vowel of the pronominal prefix coalesces. The distant demonstrative consists of the pronominal prefix plus the stem la. It is the latter that provides the evidence for the underlying low tone of the pronominal prefix of classes 1, 4 and 9. The demonstratives are presented in Table 7.4 and Table 7.5. this that that (far) yá ká hú Table 7.4: Inner Mbugu demonstratives

ADJECTIVES AND OTHER NOMINAL MODIFIERS

class 1 2

this that that (far) úu úo ulá áva ávo válá úu úo úlá 4 íi iyo ilá 5 íli ílo lilá 6 áma áo ála 7 ích i icho chíla 8 ívi ívyo vílá íi 9 ilá íyo 10 jílá ijo iji úlo lúlá 11 úlu 12 áka áko kálá 14.1 úvu úvo vúlá 14.2 úu úo úlá úku úko kúlá 15 áha áho hálá 16 Table 7.5: Normal Mbugu

181

pronominal prefix  vá ú i li á/má chi vi i ji lú ká vú ú kú há demonstratives

The demonstratives can be used attributively, predicatively and independently (i.e. with no referent present). In Inner Mbugu the demonstrative hú is used as translation equivalent of a third person pronoun. The Normal Mbugu equivalent is the pronoun yé' in the singular; for the plural, Normal Mbugu uses the demonstrative válá 'those (cl. 2), they' or the demonstrative vó without the initial a; the latter is also used in Inner Mbugu as an alternative to hú to indicate plurality. The Normal Mbugu locative de­ monstratives are used where English would use locative adverbs such as 'here', 'there'. Inner Mbugu has adverbs as equivalents áta 'here', i'í 'here', twai 'there', íla'í 'there' (see section 8.2). 7.5

Quantifiers The following NMb quantifiers use a pronominal prefix that agrees with their head noun: -o-óse or -ó-ose, 'any' [I+NMb], (i)ngi 'other' [NMb], -ose 'all', and the modifying question word -hi 'which' [NMb] (see Table 7.6). Note that the Normal Mbugu word for 'other' falls into the morphological category of quantifiers; the con­ cept 'other' presupposes partitioning just like quantifiers proper. The Inner Mbugu equivalents of the latter three are adjectives (see section 7.1), that is to say they take the nominal prefix for agreement. Thus, there is a remarkable difference between IMb and NMb in the agreement behaviour of pronouns; in this case it concerns the form or the type of agreement morpheme that is used with quantifying modifiers. The forms of these quantifiers are given in the following table. The forms for 'which' and 'other' show the underlying tones of the pronominal prefix. The quantifier -ose 'all' has a su­ perimposed tone pattern LL and o-óse 'any' shows tonal variation, that is, the tonal

182

CHAPTER 7

patterns LHL and HLL are both acceptable; the final vowel of o-óse is sometimes dropped, -ose 'all' can only be used with plurals and thus does not appear in the singu­ lar classes.4 class 1 2 3 4 5 6 7 8 9 10 11 12 14.1 14.2 15 16

Pronominal -hi 'which?' -ngí 'other' -ose 'all' -o-ós(e) or -ó-ose 'any' prefix [NMb] [I+NMb] [NMb] [NMb] uhí ungí wowóse ù vá vahí vángí vose vovóse  úhí úngí wowóse í ihi ingí yose yoyóse lí lihi Ungí lolóse á ahí ángí ose oóse hí chíhí chíngí chochóse vi víhí víngí vyose yovyóse i ihí ingí yoyóse jihí jose jojóse ji jingi lú lúhí lúngí lolóse ká káhí kángí kokóse vú vúhí vúngí vose vovóse ú úhí úngí wowóse kú kúhí kúngí kokóse há háhí hose Table 7.6: Quantifiers

8

INVARIABLES OR OTHER WORDS

After discussing verbs, nouns, adjectives, quantifiers, and pronouns, we have now come to the remainder of the word classes. Whereas the word classes discussed so far can be defined on morphological grounds, this is not so clear for the rest. The only positive morphological characterisation that can be given to all remaining words is that they are invariable. The copulas ni and si belong to this category and have already been dealt with in 5.5. The remaining words are quite diverse in function and this chapter is organised according to function rather than form. This functional organisa­ tion enables us to include words such as time, place and manner expressions including lexicalised phrases for which it is difficult to establish whether they are nouns or not because they do not easily appear in a position where they govern agreement. There is one basic division that can be made and that is between words that cannot form a phrase in themselves, and those that can. In the first category we find words that trans­ late as prepositions. The monosyllabic invariables in this category may cliticise to the following word (their complement) and undergo vowel coalescence rules. In the sec­ ond category we find nouns that translate as conjunctions such as (mi)hambu 'be­ cause', kwato 'thus', mira 'but', adverb-like invariables such as  'only', pere 'al­ ready', and words for time and place. 8.1

Dependent invariables: Prepositions Three of these prepositions have a wide range of functions. All three, na, ni and kwa, are common to Normal and Inner Mbugu. The preposition na is used for a wide range of functions: comitative ((1)), connective 'and' ((3)), additional 'and' in numeric expressions, e.g., ikumi na mnane '18'. It is also used to indicate "as well" ((4)), or to introduce a topic, "regarding, concerning" ((5)). It is never used for instrumental 'with'. The complement of na is either a noun phrase ((3)) or a relative pronoun ((1), (4)). The preposition may clitisise to a preced­ ing class prefix, as in va-na-o 2-with-REL in ((2)). (8.1) tu-ka-díndika na-vó

(NMb)

1PL-CSEC-run with-2:REL

'We ran away with them.' (8.2) u-ku-súwa wé u-tágh-w-a na va-ná-o 2SG-COND-refuse you 2SG-throw-PAS-F with 2-with-REL 'If you refuse, you will be thrown away with your fellow men.'

(NMb)

184 (8.3)

CHAPTER 8 u-lá m-borá na m-bwánge úu ve-kelaáy-e mhpombe 1-DEM3 1 -girl with 1-boy l:DEM12:PST-pass:at-PF water 'That girl and this boy passed by some water.' (NMb)

(8.4) vavee -ye áa-fwá father with-REL:l l:PST-die 'And his father died as well.'

(NMb)

(8.5) na ngasu ku-yoa mhpalé with lineage 15-go Pare 'Regarding the lineages, it was going to Pare.'

(NMb)

In addition, na is used to encode locative expressions ((6)), and temporal expressions such as na amá (IMb) or na kio (NMb) 'at night', as in (7). (8.6) tú-tóng-e na káa IPL-go-SBJ with home 'Let's go home.'

(NMb)

(8.7) he-vúka mbuá na kió 16-get:up rain with night 'The rain started during the night.'

(NMb)

Prepositional phrases with na also occur in two specific constructions: as comple­ ments of verbs containing the reciprocal extension -an, regardless of whether the re­ ciprocal is lexicalised or not ((8)); and in the subject inflected preposition na to indi­ cate, among other things, possession ((9)) (see section 5.5). (8.8) ji-pátana na ki-zighúla 10-answer with 7-Zigua 'They answered in Zigua.' (8.9) e-ná mi-síku, mi-zinga ya 1-with 4-beehives 4-beehives(Sw)4:of 'He had beehives, beehives for honey.'

(NMb)

ndóki honey

(NMb)

The preposition ni introduces the actor ((10)) or, in combination with the question word "what?", the cause ((11)). (8.10) vaMa'á te-vé-áhó-ik-a ni vahé ve-ló maána Mbugu NEG-2-see-ST-F by people 2-have meaning 'The Mbugu are not seen by people as being worth a thing.'

(IMb)

INVARIABLES AND OTHER WORDS (8.11)ni ahhoní é-zó by what 1 -cry 'Why does he cry?'

185 (IMb)

In the construction with a stative extension in ((10)) above, the preposition ni is to in­ dicate the actor; the most frequent usage of ni is to indicate the actor in passive sen­ tences ((12)). It is also used with verbs that contain a frozen passive ((13), (14)). Simi­ larly, it may be used in loans from Swahili that contain a passive ((15)). However, ni may also be used in passive sentences, not to indicate the actor (or subject of the cor­ responding non-passive sentence), but to indicate an argument that is neither the actor nor the instrument, but a crucial factor in the process ((16), (17)). (8.12)tw-áa-odoshwa ni mbúa lPL-PST-get:wet:PS by rain 'We have become wet because, of the rain.'

(NMb)

(8.13)é-hláwe ni mzunguré. l-fall:ill:PS:PF by back 'His back aches/

(IMb)

(8.14)u-sánuiwe ni 2SG-be:angry:PS:PF by 'Why are you angry?'

ahoni

(IMb)

what

(8.15)á-shtakiwa kángi ni va-vugha. l:PST-be:accused(Sw) again by 2-Vugha 'He was accused again by the people of Vugha.'

(NMb)

(8.16)iki m-bwángi e-vwír-w-a now 1-boy 1-anoint-PAS-F 'The boy was anointed with oil.'

(NMb)

(8.16)hé-rúgh-w-a ni 16-cook-PAS-F by plates 'It is prepared on plates.'

ni by

ma-futa, 6-oil nyingó

(NMb)

The preposition kwa, which I have glossed in the examples as 'with' indicates instru­ ment ((18), (19)), manner ((20), (21), (22)), field of comparison, measurement, judgement ((23), (24)), or in general "regarding" ((25)), as well as locative source ((26), (27)). Kwa is also used in the expression wé kwa wé 'at once' (lit. one with one). The preposition kwa may clitic ise to the following word, as in kwézoka from kwá izoka 'woth an axe' in (19). This preposition kwa is not related to the noun kwá (class 9) meaning 'path' in Inner Mbugu since the noun is a truncated form of kwáhli and

186

CHAPTER 8

thus presumably of Non-Bantu origin while the preposition kwa is common to Pare, Shambaa and other Bantu languages. (8.18)áa-té mihatú kwa choká' 1 :PST-cut trees with axe 'He has cut the trees with an axe.'

(IMb)

(8.19)áa-tema miti kwézoka l:PST-cut trees with:axe 'He has cut the trees with an axe.'

(NMb)

(8.20)mhpého í-hunhta kwá ngavu wind 9-blow with power 'The wind blows with force.'

(NMb)

(8 .21) ni-úfú mwáhla kwá ú 1SG-blow fire with force 'Í blow the fire with force.'

(IMb)

(8.22)

ni-he-súngíríja kwá anháoni-hó ni-áho ní shuhé. 1SG-PST.COND-turn:around with care COP-16:REL 1SG-see COP leopard 'When I turned around carefully it was that I saw it was a leopard.' (IMb)

(8.23) nhtembo i-táho mboghó kwa ugirú elephant 9-pass buffalo with bigness his 'The elephant is bigger than the buffalo.'

kú'u

(IMb)

(8.24) ingcnva ni m-tana kwa súra mirá t-é-ghuh-í-w-e (NMb) although COP 1-beautiful with face but NEG-1-marry-APL-PAS-PF 'Although she has a nice face, she is not married.' (8.25) ku-dóka nguo kwa mwaána ni 15-dress clothes with child is 'To dress a child is a difficult job.'

ndimá work

m-bahá 9-big

(NMb)

(8.26) lungorí lu-vónekana kwa haé hill 11-visible with far 'The mountain is visible from afar.'

(NMb)

(8.27) hú ighámba lí-aho-ik-a kwa haé that rock 5-see-ST-F with far 'The mountain is visible from afar.'

(IMb)

187

INVARIABLES AND OTHER WORDS

The preposition kwa is used in a number of fixed expressions introducing reason com­ plements. Parallels can be found in Swahili expressions for 'because' such as kwa sababu, kwa ajili ya, kwa jinsi (ya), etc. expression kwa mburi ya kwa hi kwa kubá kwa chila kwa gigi kwa kini Table 8.1:

language source meaning example # IMb mburi matter 28 NMb iti reason 29 IMb ku-bá 15-say 30 NMb chilá 7:DEM3 ' that thing- 31 IMb cause 32 gigi NMb kini what? kwa in fixed expression meaning 'because, why'

(8.28)áa-rónya nyamálo kwa mburi-y-a 3SG:PST-leavework with reason-9-CON 'He stopped the work because of the rain.'

mare rain

(IMb)

(8.29) kwa hi w-á-cheléwa si-ka-tet-e na wé (NMb) with reason 2SG-PST-late lSG:NEG-NEG:FUT-talk-NEG with you 'Because you are late, I won't talk to you.' (8.30) kwá kubá w-áa-hléiwa sí-ka-yó na ári with saying 2SG-PST-late 1SG:NEG-NEG:FUT-speak with you 'Because you are late, 1 won't talk to you.'

(IMb)

(8.31)é-ku-íngia na vúgha kwá chí-la ku-síikia ha-la va-kirindí(NMb) 1-COND-enter with Vugha with 7-DEM3 15-listen 16-DEM3 2-Kilindi 'He went to Vugha because he had heard it was the place of the Kilindi.' (8.32) kwa gigi w-áa-hléti si-ka-yó na ári with cause 2SG-PST-late 1SG:NEG-NEG:FUT-speak with you 'Because you were late, I won't talk to you.'

(IMb)

(8.33) kwa kini a-ku-hangwa nhkuku nyéwa with what 3SG-COND-call:PAS chicken black 'Why is he called the Black Chicken?'

(NMb)

The preposition sa: The invariable sa 'like' occurs in both languages and is also pre­ sent in Shambaa. Its semantic range goes from the expression of similarity ((34)) and exemplification ((35)) to comparison ((36), (37)), and from characterising ((38)) to that of pointing ((39)). Functionally sa can modify a noun ((35)), a noun phrase ((34)), or introduce spatial adverbials ((39)).

188

CHAPTER 8

(8.34) é-zá-dóri gunda sa hú lá-kwé náaní. l-IT-take:out pipe like that 5:his whatyemacallit 'He took a pipe like that of you know who'

(IMb)

(8.35) ni sa vigí, ni makeréno is like things is earrings 'It is like things, like earrings.'

(IMb)

(8.36) tu-ka-jinhka na-hé sa waivi ú-ku-jinhka na mché wako. 1PL-CSEC-run with-16 like now 2SG-COND-run with wife l:your 'and we ran away like nowadays, if you run with your wife.' (NMb) (8.37) kigi yá ni ki-helí sa thing DEM1 is 7-hard like 'This thing is as hard as iron.'

chuma iron

(IMb)

(8.38)e-kúnda hambú véghuh-íye sa mkalá. 1-want because 2-take:away-PF like hunter 'He wanted (it) because they took (him) as a hunter.'

(NMb)

(8.39) íki tii kawaida mhaká ni now reason normal(Sw) border is 'It is normally up to here.'

(NMb)

sa like here

áho.

There are a number of words that are very similar in use to the prepositional invari­ ables discussed so far. Some of them are borrowed from Swahili, such as m 'un­ til' (I+NMb), kila 'every' (I+NMb), kuliko 'than' (I+NMb). Others are actually verbs such as kuli (IMb) and kufuma (NMb) '(come) from'. In other cases their syntactic behaviour is ambivalent as a consequence of an on-going process of grammaticalisation. This is the case for the class 16 locative noun prefix he (I+NMb) and the class 7 genitive pronoun chá (I+NMb). For each of these I give a few examples below. The preposition má is a borrowing from Swahili  where it is a noun meaning 'border' which can be used as the functional equivalent of a preposition meaning 'un­ til, up to'. It is in this usage that the word was borrowed. For the nominal function Normal Mbugu has the corresponding class 9 noun mhpaká (I+NMb) 'border' with the regular voiceless pre-nasalised stop. Many languages of Tanzania use this Swahili bor­ rowing for the prepositional function given above. The complement of the preposition indicates place, e.g., kuli kinhko mpaká maghambá 'from Kinko to Magamba'(IMb) (see also (40)). It is also used in a temporal sense, e.g., má chamageri 'until the evening' (IMb) (see example (41)); and it is used for a telic event ((42)), or for the endpoint of an enumeration ((43)). The noun mhpaka 'border' can also be used to mean 'last', i.e. the one up until now ((44)).

189

INVARIABLES AND OTHER WORDS (8.40) ve-ojíj-w-e má mpakani. 2-send-PAS-PF until(Sw) border:in(Sw) 'They were sent to the border.'

(NMb)

(8.41) mpaká hódi m-né-gáya until place 2PL-FUT-die-APL-F 'until the time that you die.'

(IMb)

(8.42) mpaká e-ka-vá-shínda until l-CSEC-2-win 'until he overwon them.'

(I+NMb)

(8.43) kufúmia tarafá y-a Mlóla, y-a Míalo 15-come:from:APL ward(Sw) 9-CON Mlola 9-CON Mlalo j-íngi úko. kuendelea kadhalíka mpaká wiláya 15-continue other(Sw) until(Sw) district(Sw) 10-other there 'coming from Mlola, Mlalo, etc up to other districts.'

(NMb)

(8.44) mwáka úo ni mh(p)jaka w-ó-manj-e ni year 3:DEM2 COP border 2SG-REL-know-F is 'The last year you know is like this.'

kama sá ao like(Sw) like 6:DEM2 (NMb)

The prepositions kulí (IMb) or kufuma (NMb) are in fact the infinitive (i.e. class 15) verbs 'come from' and can be used to introduce a locational noun to render the "source" meaning 'from', as in kuli hakánu 'from us' (IMb), and in (45). (8.45)iti ve-ghuh-iye kufuma h-ákwe reason 2-take:away-PF from 16-his '...because they took it from that Mrutu'

Mrútu Mrutu

(NMb)

The preposition kila: The pre-nominal modifier kíla or kira meaning 'every' is bor­ rowed from Swahili or Shambaa, where it is also common, as it is in most languages of Tanzania. Examples are kila mhé kikú'u 'every person his (=what belongs to him).' (IMb), and kira msi 'every day' (NMb). The preposition hé: The locative class 16 prefix often precedes the noun prefix of a noun as in (46) (see also section 6.4). However, since it can be followed by pronouns ((47)) and by demonstratives ((48)), it is much less fixed to the noun than other noun prefixes that precede a noun class prefix, such as class 6 ma in ma-lu-go 6-11-fence 'fences' (I+NMb), see section 6.3, where no material can intervene between the two noun prefixes. Numerous place-names begin with hé.

190

CHAPTER 8 (NMb)

(8.46)é-tónga hé-nyumbá 1-go 16-9:house 'He goes home/

(8.47) va-ka-ni-pa'á-y-a haghó hé wa yá i'enéoyá l-á shúre 2-CSEC-1SG-cut-APL-F 16-my 16 3-CON this area this 5-CON school 'They measured for me my place at this area of the school.' (IMb) (8.48)

tw-ée-ikée áha hé ili isangá lPL-PST-stay:PF here 16 5:DEMl land 'We lived here in this land.'

(NMb)

The preposition chá : The genitive pronoun for' class 7 chá can refer to something very general as 'thing' as in (kitu) cha kuendelea 'something to continue', and it can be used to indicate manner as in ch-avo 7-their 'their way (manner)'. Furthermore it is added to words of time and place such as those given in Table 8.2. In this latter usage it behaves like a preposition. expression meaning chábosí 'at first' chainsí 'today' cháávo 'in the morning' cháhlen 'in the morning' cháléla 'South' cháŋila 'North' Table 8.2: Expressions with

meaning source bosi 'first' msi 'day' yavo 'tomorrow' (NMb) hlenu 'tomorrow' (NMb) lela 'under' aŋila 'above' (IMb) general locative use of chá

The preposition kuliko is a loan from Swahili which was only attested in sentences directly translated from Swahili containing a comparative or superlative ((49), (50)). The preferred means of expressing comparative and superlative meaning is with the verb "surpass". (8.49)nhtembo ni girú kuliko nihéna kabúna elephant is big than animal(s) all 'The elephant is the biggest of all animals.'

(IMb)

(8.50)nhtembo ni m-bahákuliko vy-óse elephant is 1-big than 8-all 'Elephant is the biggest of all (things).'

(NMb)

INVARIABLES AND OTHER WORDS

191

8.2

Independent invariables The independent invariables are those words that can form a phrase in them­ selves and that cannot be shown to belong to any of the other word categories, i.e. nouns, verbs, adjectives, pronouns. In their function they resemble adverbs. The or­ ganisation of the presentation of these invariables below is according to semantic do­ mains: time, place, manner, other.

Expressions for time: From "today", the Mbugu count two days back and two days ahead in both languages, see Table 8.3. All the forms are simple roots, i.e. they are not morphologically complex. Instead of yó 'today' (NMb), one can also use chá-msi from msi (3) 'day' (NMb). 'Tomorrow' is also used in the meaning of 'the next day' ((51), (52)) and to form expressions like mkó wa hlenú (IMb), mw-áka w-á yávo (NMb) 3year 3-of tomorrow 'next year'. The words from Table 8.3 may be considered nouns since they can occur in copular constructions ((53), (54)). form wásuhú izána ósa ighuo wa'í yó hlenú yávo mtóndo húgo

meaning language day before yesterday IMb day before yesterday NMb yesterday IMb yesterday NMb today IMb today NMb tomorrow IMb tomorrow NMb day after tomorrow NMb day after tomorrow IMb Table 8.3: Relative days

(8.51)tó hlenú vé-jí-géra sa sita there tomorrow 2-10-bring hour six 'The next day they bring the cattle at twelve o'clock.'

(IMb)

(8.52) ha-lá yavó e-kúnda -za-kasha 16-DEM3 tomorrow I-want 15-IT-hunt 'The next day he wanted to go hunting.'

(NMb)

(8.53) hlenú ni m'azé mína tomorrow COP day how:many 'What day is it tomorrow?'

(IMb)

192

(8.54)

CHAPTER 8 ighuó ni kázi ka-ngáhi yesterday COP work:day time-how:many 'What day is it tomorrow?'

(NMb)

(Work)days of the week are from the Swahili word kazi 'work', plus a numeral count­ ing from the first working day, Monday: kazi bosi (IMb) 'Monday'; kázi ka nú (IMb) 'Tuesday', kázi ka xaí (IMb) 'Wednesday', kázi ka háí (IMb) 'Thursday', kázi ka koí (IMb) 'Friday'. The days of the weekend are non-derived forms: mkalagázo 'Satur­ day', jumá 'Sunday' (cf. Sw. ijumaa 'Friday'). Most Mbugu people are Christians and therefore the Swahili and Muslim special day of the week, Friday, is equated with the Christian special day of the week, Sunday. The periods of the day are given in Table 8.4. Some of them require the preposition chá, see 8.1. There are several expressions that translate as 'now', ingei, ivi, waiví, iki (NMb) and iji, wa'i iji (IMb). Another time expression is kaé 'long ago, past' (I+NMb). form meaning cháhlenú morning chá(he)ávo morning chámagherí midday ama night kió night xwa'ú last night amá'amá early morning, sunrise Table 8.4: Parts of the day

language IMb NMb I+NMb IMb NMb IMb IMb

Locative expressions: Some locative expressions are given in Table 8.5. These include general words for 'place', adverbials for 'here'and 'there' and other directions and positions. Some of these are derived from body parts, such as lia (IMb) 'buttocks' for 'behind'; some are nouns with a non-locative basic meaning, such as kaa (NMb) 'sky'for 'high'. form hódi hanhtu aré, arái áta aha, áho Vi ila'í k

meaning place place place, there here here here here here

language IMb NMb IMb IMb NMb IMb IMb NMb

INVARIABLES AND OTHER WORDS

form to twáy hasi ghati ughu aŋila di sí lela ishí chálela kusini cháŋila kaskazin i lía nguma naré kará mbái hombáy zemera (ki)bwa nhkade sélá oju gána shigati úda haé siná ana kaa

meaning there, thus there middle middle above above down down under lower south south north north backwards, behind, buttocks backwards forwards sidewards sidewards front front side, part side other side other side outside outside far far near sky, high sky, high Table 8.5: Locative expressions

193

language IMb NMb IMb NMb NMb IMb IMb NMb IMb NMb IMb NMb IMb NMb IMb NMb IMb IMb NMb NMb IMb IMb NMb NMb IMb IMb NMb IMb NMb I+NMb IMb NMb

Expression of manner: In Table 8.6 the manner adverbials are given. Those of speed — 'quickly' and 'slowly* — are reduplicated; the same applies to the word for 'care­ fully'.

194

CHAPTER 8

form wá kángí

fyafya

meaning again again, thus quickly quickly slowly slowly carefully

language IMb NMb NMb IMb IMb NMb IMb

wanguwángu hluhlu mhpóamhpóa anhkáoanhkáo together hawé I+NMb first I+NMb bosí Table 8.6: Manner expressions

The invariables in Table 8.7 are conjunction-like. They occur in sentence initial or second position unless otherwise stated. form meaning language remarks so, reason íti IMb thus IMb kwato mirá but IMb ila but NMb mbánye well I+NMb angái if IMb nanga if NMb ité that IMb + optative néri or, no IMb or óo NMb kubá that IMb from 'to say' hángi maybe I+NMb kweri truly NMb úgí truly IMb du only NMb clause-final bi only IMb clause-final pere completely I+NMb clause-final Table 8.7: Conjunction-like invariables 8.3

Independent personal pronouns The independent personal pronouns are used for emphasis only. Usually subject concord on the verb is sufficient for the expression of arguments of the clause. Both the Inner Mbugu third person singular and thė Normal Mbugu third person plural are identical to the demonstrative meaning 'that (far away)', hú in IMb and válá (class 2) in NMb. The class 2 distal demonstrative vó (without the initial a, see section 7.4) can

INVARIABLES AND OTHER WORDS

195

also be used as a third person plural pronoun in both Inner and Normal Mbugu. The personal pronouns of Normal Mbugu are identical to those in Pare. Inner Mbugu Normal Mbugu mmí ání arí wé (h)ú yé nine, nne úswi kúné únywi (h)ú válá Table 8.8: Independent personal

meaning I you she/he we you they pronouns

Personal pronouns in Inner Mbugu have the same distinctions as those of Normal Mbugu, see 3.9. 8.4

Higher numbers and other invariable modifiers The lower numbers are adjectives (see section 7.1). form mtandàtu matezo sitá mfungáte sabá mnáne kendá ikádo ikúmi igana mía élfu or álfu

meaning 6 6 6 7 7 8 9 10 10 100 100 1000

language I+NMb IMb I+NMb I+NMb I+NMb I+NMb I+NMb IMb I+NMb I+NMb I+NMb I+NMb

remarks unusual rare from Sw unusual from Sw

from Sw from Sw

Table 8.9: The invariable numerals For those numbers for which various forms are used, those that are closest to Swahili, i.e., sitá for 6, saba for 7, ikúmi for 10 and mia for 100, are more commonly used. For 1000, there is only a Swahili loan. Non-basic numbers are formed by addition marked with na and multiplication marked by juxtaposition in head-modifier order. The added lower numbers that are adjectives need an agreeing noun class prefix, also when they follow na in higher numbers (indicated with — in Table 8.10).

196

CHAPTER 8

form meaning language IMb ikádo na wé 11 11 NMb ikúmi na —mwé 12 IMb ikádo na nú 12 ikúmi na —iri NMb milongo minú 20 IMb ikádo ka nú 20 IMb milongo minú na kái 23 IMb ikádo ká ikádo 100 IMb mia nú 200 I+NMb magana mairi NMb 200 Table 8.10: Compound numbers

9

NOTES ON SYNTAX, CODE-SWITCHING AND TEXTS

9.1

Syntax This section contains a few remarks on the syntax of Mbugu and does not aim at a full syntactic analysis. The core of the sentence is the verbal phrase, i.e. the inflected verb including clitics and the object noun phrase. The subject is obligatorily indicated on the verb and if a lexical subject is present in the sentence then its normal position is before the verb. The object follows the verb and can, but need not, be represented by a prefix on the verb. Any other phrases are not part of the frame of the verb but may represent a prominent participant of the happening or event, e.g., the agent in passive and thus object-less sentences ((1), (2), (3)). (9.1) é-ka-fw-í-wa ni m-bwange 1 -CSEC-die-APL-PAS by boy 'She was met with death of a boy (i.e., the boy died).'

(NMb)

(9.2) tu-ga'á-í-we churú ni babá (IMb) 1PL-kill-APL-PAS bull by father 'We were slaughtered a bull by father (i.e. father slaughtered a bull for us).'1 (9.3) ji-tang-wa futu ni vashamba 10-call-PAS grasshopper(Sh) by shambaa 'They are called "futu" by the Shambaa.'

(NMb)

Apart from the verb phrase and the noun phrase, another types of phrases are the prepositional phrase (PP), and the adverbial phrase (AdvP), as in (4) which has the structure: [ppkwa kubá [ppkulí kinhko][ppmpaká magamba]] ni [npmuda [npwa madamo mahai]]] [advpto]. (9.4) kwa kubá kuli kinhko  magamba ni muda wa madamo with saying from Kinko until Magamba COP time 3:CON hours mahai to (IMb) 6-four only 'That (the distance) from Kinko to Magamba is a period of four hours only' Locative complements consisting of an invariable phrase with na are used with verbs such as fuma 'come from' (NMb), samia 'pass' (NMb), yoa 'go to' (NMb), ingia 'en­ ter' (NMb), za 'come' (NMb), kela 'pass', tonga 'go' (NMb), hunduwa 'be returned' (NMb), hunduka 'return' (NMb), gwa 'fall' (NMb), dindika 'run" (NMb). Sentence (5) shows that the invariable na is still added when the speaker cannot think of the word for the locative referent.

198

CHAPTER 9

(9.5) tukaíngia na nóoni 1PL-CSEC-enter with whatyemacallit 'We entered eh...'

(NMb)

9.2

Set sentences In this section I give the most frequent set expressions such as greetings and ad­ dressing expressions. The sequences of set utterances that are used when two people meet and exchange greetings are represented in Table 9.1 and Table 9.2. greeting mwaáká 1Q 1A twáaka sarí kúne 2Q 2A1 ní mkuhló 2A2 twáaka mwaáká kikuhló 3Q  twáaka kikuhló 4Q yóní kiamá 4A ni kikuhló sari kikághu 4Q 4Q' yóní ma húndo 4Q' sarí makághu 4A' ni makuhló yó (ni) m'azé 5Q 5A ni mkuhló sari mkágho 5Q 6Q yóni chakakó Table

gloss 2PL:PST:get.up lPL:PST:get.up news 2PL COP l:good lPL:PST:get.up 2.PL:PST:get.up 7:good lPL:PST:get.up 7:good today:COP 7:night COP 7:good

function opening

news 7:2PL.POSS

reversal

greeting mwaávúka twaávúka sari únywí twaávúka mwaávúka kitana twaávúka kitana ghambáni msí ní mtana sari wényú Table

gloss 2.PL:PST:get:up lPL:PST:get:up news 2PL lPL:PST:get:up 2PL:PST:get:up 7 good lPL:PST:get:up7: good ?:COP day COP 1 :good

1Q 1A 2Q 2A 3Q  4Q 4A 4Q

asking for news second inquiry wishing good morning

today:COP 6:morning news 6:2PL.POSS

COP 6:good today(:COP) (mid)day COP l:good

wish uttered during the day

news 1 :2PL.POSS

today:COP at:? 9.1: Sequence of greetings (IMb) function opening asking for news reversal wish uttered during the day

news 1 :2PL.P0SS

9.2: Sequence of greetings (NMb)

Requests are accompanied by the words namálále or naiishe, translatable as 'please'

SYNTAX, CODE-SWITCHING AND TEXTS

199

One asks for forgiveness by using mbwangi. There are several words for thanking, namely sara'é in the Rangwi variety of IMb, héika in the Magamba variety of IMb, yahálo in archaic IMb and hongé(r)a in NMb. A call for attention over distance is wérewé; a call for attention when handing over something is hĩ(I+NMb), or the imperatives dóri 'take out' (IMb) or géra 'put in' (NMb). The following calls to summon animals are common to both languages: dog fiífií (or whistle) chicken kúk kú cat mtsmtsmts cow ẽẽ ẽẽ goat mẽẽ mẽẽ sheep bãã bãã Both languages use the filler nóoni 'thingummy, whatyemacallit'. 9.3

Code-switching Code-switching does occur between (stretches of) Inner Mbugu and Normal Mbugu. Within such stretches intrusions of words from the other parallel lexicon and from Swahili do occur. Both kind of switches (inter and intra-sentential) are easily recognisable but both types of code-switches are problematic for the Matrix Language Frame model of code-switching put forward by Myers-Scotton (1993a). The reason is that the switch between Inner Mbugu and Normal Mbugu cannot be recognised as a switch in that model because there is no change in system elements: grammatical, in­ flectional morphemes remain from the same language. Consider sentence (9.6): The intrusion seems to be the IMb possessive kanu. The first phrase is a name and not dis­ tinguishable as being either Inner Mbugu, Normal Mbugu or Swahili; the second phrase contains a Swahili verb root omb but the inflectional and derivational affixes could be either Swahili or (I+N)Mbugu; the last phrase contains a preposition that is common to I+NMb, the word 'father' which is also common to both languages, and an IMb possessive. The latter is quantifcational and thus a system morpheme in MyersScotton's terms and that would make Inner Mbugu the Matrix Language in which case the affixes in the second phrase also count as IMb, as well as everything else that is common to both languages. The only embedded element would be the Swahili verb root and the intrusion of the possessive is not recognisable as such in a strict interpre­ tation of the MLF model. In this and following examples Swahili intrusions are under­ lined and Inner Mbugu elements are in bold. Normal Mbugu elements for which there is an Mb equivalent are in bold. (9.6) [maja Sendresi] [a-ka-omb-w-a] [ní [abá kanu]] Major Sanders 1-CSEC-request-PAS-F by father our 'To Major Sanders a request was made by our father.' However, Inner Mbugu or Ma'á by itself (with system morphemes from one source

200

CHAPTER 9

and content morphemes from different sources) can be described and understood in terms of the Matrix Language Frame model (or similar code-switching models). Ma'á is then a special instance of a Matrix Language Frame with the prerequisite that it takes every content morpheme from the embedded lexicon. Nonetheless the pronouns, possessives, and demonstratives have to be analysed as content morphemes and not as system morphemes (in the MLF model would these be system morphemes). Muysken (1997) distinguishes three patterns of code-switching: alternation, in­ sertion, and congruent lexicalization. Alternation is exemplified by the switch from Normal to Inner Mbugu in the sample below. Insertions are the Swahili embeddings and the NMb intrusions into IMb (and vice versa). Congruent lexicalization "refers to a situation where the two languages share a grammatical structure which can be filled lexically with elements from either language." (Muysken 1997:362). Though different from the examples that he mentions, according to his definition, the relationship be­ tween Inner and Normal Mbugu can be interpreted as complete congruent lexicaliza­ tion (within which alternation occurs); within the alternating stretches limited insertion can take place. Some of the properties that Muysken attaches to congruent lexicaliza­ tion are valid for Mbugu. These are selectivity (the switched fragment is selected by an element in the other language), linear equivalence, function words in a switched fragment, phonological and syntactic adaptation. Other properties of Mbugu as a form of code-switching van bee seen as indicative of the insertion type of code-switching (single constituent, a-b-a nested structure, nouns or adjectives as inserted elements) but this again is a consequence of the fact that if IMb speech is taken as codeswitching, such code-switching is applied maximally (namely for each root). Conclud­ ing, models for code-switching can account for the structure of Inner Mbugu (Ma'á) with its dichotomy of content versus system morphemes, but only if these models are generalised to cover all instances of multiple language use. The peculiar, radical char­ acter of Inner Mbugu (when viewed as an instance of code-switching) and the fact that within Inner Mbugu switching of codes is still possible and recognisable, renders the strict application of code-switching models meaningless. Still, the insights from the code-switching literature also hold for code-switching within Inner and Normal Mbugu; for example, pragmatic sentence introducers and conjunctions favour a switch to the other language. The intrusion of Normal Mbugu personal pronouns and posses­ sives into Inner Mbugu stands out as a peculiarity of Mbugu code-switching. I take it to be an indication that although parallel IMb personal pronouns and possessives do exist and are used, this domain of the parallel lexicon is most likely to be considered as consisting of system morphemes rather than content morphemes. 9.4

A sample of code-switching The following is an excerpt from an interview by my research assistant William Mboko (W) with Mzee Ishika (I) in Malibwi on 10.10.1993. The subject of this part of the interview is the fact that the Mbugu have two languages. The elements that are markedly Normal Mbugu are in bold; those that are markedly Inner Mbugu are in ital­ ics; the Swahili elements are underlined.

SYNTAX, CODE-SWITCHING AND TEXTS W: i-zina 1-áko u-táng-w-a mzee Ishíka Salim Mvewa 5-name 5-your 2SG-call-PAS-F mister Ishika Salim Mvewa 'You are called Mister Ishika Salim Mvewa?' W: u-na mi-áka mí-ngahi mpaká waívi 2SG-with 4-year 4-how:many until now 'How old are vou ? (How many years do you have up to now?)' I: ná-ku-ti ni-na mi-áka sabini na mi-irí mpaká waívi lSG:1NC-2SG-say 1SG-with 4-year seventy with 4-two until now 'I am seventy two (I tell you I have seventy-two years up to now.)' W: ingéi í-chi ki-mbughu chó chá aina mbirí now 7-7 7-Mbugu 7:REL 7:CON sort two 'Well this Mbugu language is of two sorts.' I: eeyé í-chi chó tu-tet-íye chá wanga yes 7-7 7:REL IPL-talk-PF 7:CON high 'Yes, this that we talk, of high../ W: na í-chi chá ndeni with 7-7 7:CON inside 'and this of inside.' W: chee-ok-íye kizéze 7:PST-become-PF how 'How did it come about?' I: chée-ok-íye va-mbúghu vé-téta chá-vó 7:PST-become-PF 2-Mbugu 2-say 7:CON-them 'It was that the Mbugu spoke theirs.' I: í-chi í-chi í-chi 7-7 7-7 7-7 'This, this, this.' I: úswí tu-ghuh-iye chá ki-asu we 1PL-take-PF 7:CON 7-Pare 'We took the one of Pare.' I:

201

202

CHAPTER 9

ambú tw-ée-tú-e na va-asú sana thus 1PL-PST-live-PF with 2-Pare very 'Because we lived a lot with the Pare.' I: basi ingei tu-ka-íkaa sa tu-ghuh-a ch-ásu sana mirá well now 1PL-CSEC-live as IPL-take-F 7-Pare very but ki-mbughu ch-étu tu-chí-gur-ísh-e sana ivi chi-tésh-w-a 7-Mbugu 7-our 1PL-7-hold-INT-PF very now 7-loose-PAS-F ní í-chi ki-swahéri by 7-7 7-swahili 'Well we lived and took Pare a lot, but ours, we hold it tight, it is being lost by Swa­ hili.' I: íngei u-kú-ta-aho-i-a mpaká waívi now 2SG-COND-EVI-see-APL-F until now 'Until when?, do you think/ W: va-mbughu vá-ngi vá-tek-i-a hé í-chi ki-shamba 2-Mbugu 2-many 2-lose-APL-F 16 7-7 7-shambaa 'Many Mbugu get lost in Shambaa.' W: ingei u-va-tégemea iti á-va va-nhtu vé-ne-za-húnduk-i-a now 2SG-2-depend:on that2-2 2-person 2-EXPEC-lT-return-APL-F ki-mbughu 7-Mbugu 'Now you depend on those people that they return to Mbugu/ I: va-mbughu vá-ngi ve-re-gur-ish-a 2-Mbugu 2-many 2-BGND-hold-INT-F mirá á-va vana vá-vo vé-ha te-vé-ka-chi-mány-e but 2-2 2:child 2-them 2-16 NEG-2-NEG:FUT-7-know-F 'Many Mbugu hold it tight but their children will not know it.' switch to Inner Mbugu I: ve-ta-hléru yaani va-ku-ta- 'áyja néesa vá-dori lugha 2-EVI-grow that:is 2-2SG-EVI-show that:is 2-take:out language y-á-vó 9-CON-them 'They show you that they give up their language.'

SYNTAX, CODE-SWITCHING AND TEXTS

203

I: ni sawa kabisa is OK really 'It is true.' W: íji mpaká wa'íjijí yá ki-Ma'á u-kú-wésía ari now until now this 7-Mbugu 2SG-COND-look:at you hé miko i-Iit-iye hi-ne-kwa kimomo 16 years 4-come-PF 7-EXPEC-be how 'Well, upto now, this Mbugu, if you observe it, you, in the years to come, how will it be?' I: hé mikó i-lit-íye chi-ta-Iaha. ní-chó íji na nne 16 years 4-come-PF 7-EVi-lost. is-7:REL now with we 'In the years to come it will get lost. It is how we are.' W: tu-dúmu tu-chi-btu hé vitábu ili kusúdi chí-sée-téka IPL-want lPL-7-put 16 books so purpose 7-NEG-lost 'We want to put it into books so that it doesn't get lost.' I: na hú ní ki-gí cha siri with that is 7-thing 7:CON secret 'and that (Mbugu language) is something of a secret.' Apart from the switch from Normal to Inner Mbugu (indicated by the —line), the ex­ cerpt shows other instances of code-switches. The most obvious ones are the Swahili elements. These include a set clause like ni sawa kabisa 'it is really true' (once), and various one word insertions. These follow the well-established code-switching pat­ terns. The Swahili words that are often used in Mbugu conversation fall into five categories: 1. Sentence initiators such as basi 'well', hata 'even', yaani 'that is', kwa mfano 'for example', kwa hio 'thus', halafu 'and then', sawa 'OK', sasa 'now', ili 'in order to'; 2. Numbers, and words like mpaká 'until' (the voiced pre-nasalised p violates Mbugu phonology), kila 'every' and sana 'very' (the I+NMb equivalent is a clitic -sha, a separate word sana provides possibilities of stress) which are nearly al­ ways from Swahili in any language of Tanzania; 3. General words (hyperonyms): aina 'manner' (no I+NMb equivalent), lugha 'lan­ guage' (IMb kiomé, NMb kitéto are both derived), siri 'secret' (I+NMb words for 'secret' refer to specific kinds of secrets), wakati 'time' (IMb madamo, NMb maróngo), shida 'problem (in its widest sense)', dawa 'medicine, any applicant' (I+NMb words for 'medicine' are more specific and are related to trees and roots),

204

CHAPTER 9

kazi 'work' (IMb nyamálo, NMb ndimd), pata 'get' (IMb aho, NMb vona), fanya 'do' (IMb sa, NMb ghendd); 4. Cultural concepts such as kitabu 'book, fundís ha 'teach', sauti 'voice' which are already loans in Swahili, just like most words in categories 1 and 3; 5. Certain terms of address, such as mzee 'old man, mister'. In addition to the Swahili insertions, there are also Inner Mbugu insertions in a Nor­ mal Mbugu stretch and vice versa. These insertions always consist of just one root (or one word). It is remarkable that I have no examples of two subsequent intrusions of Inner Mbugu roots in a Normal Mbugu stretch or vice versa. Insertions of Inner or Normal Mbugu can be categorised as follows: 1. Sentence introducers, e.g., ingei (IMb) 'now', mira (IMb) 'but', sa (IMb) 'as', iti (IMb) 'that' in the NMb part of the conversation above; 2. Quantificational function words such as ngi (NMb) 'many', ose (NMb) 'all', Kangi (NMb) 'again' are recurrent NMb intruders into IMb texts; 3. Words for which there is an appropriate IMb form but which for one reason or another is not readily available to the speaker, e.g., teka (NMb) 'loose' in the final sentences of the conversation excerpt;2 4. Words that are defined for the conversation. For example, I recorded a conversa­ tion between four people about former customs, see 9.5. In this conversation there were a number of words that are core words within this conversation (dance, step in a dance, neck, necklaces and other ornaments) that are always referred with the same word form. This is in fact common practice in conversation. In this particu­ lar case these words constitute NMb intrusions in an otherwise IMb text which is probably due to the accommodation of some participants who were less at ease in IMb. 5. Possessives and personal pronouns are often from NMb in IMb texts. In the afore­ mentioned conversation of four people on former customs, which is in IMb, all pronouns and all possessives were from NMb except one (curiously enough the least common pronoun, the second person plural). This may be an extreme case, as in the conversation cited above in extenso, there is only one unequivocal NMb intrusion, a third person singular possessive. In addition, the excerpt above has a third person plural possessive from NMb in the second IMb part (and a NMb sec­ ond person plural possessive in the first sentence following the excerpt cited here). I have mostly mentioned instances of intrusions of NMb into IMb and only a few cases of the opposite, but this may be a correlate of the fact that I have very few dia­ logues in NMb, as well as the fact that code-switching in monologues is rare. The rela­ tively frequent intrusion of NMb pronouns and possessives into IMb is remarkable and seems to me an outstanding characteristic which is uncommon in code-switching.

SYNTAX, CODE-SWITCHING AND TEXTS

205

9.5

Texts The following texts serve as examples of more or less natural language use in Mbugu. I have selected them from a wider range of taped and analysed texts. Two of the texts in this presentation are conversations and the two others are monologues (oral histories). In my experience in fieldwork among the Iraqw, Alagwa, Datoga, Bowe, Zigua and Mbugu in Tanzania, the Mbugu are a remarkable exception in that the verbal arts seem to be practically non-existent. On several occasions I showed in­ terest in taping stories, riddles, songs, etc. Songs I got, but stories and riddles I did not get a change to tape, despite several promises for stories which always turned out to be oral histories or anecdotes. Conversation on tradition (IMb, NMb) This converation was taped in Magamba in February 1993 when I visited, to­ gether with my assistent, William Mboko (W), two of his aunts, Mkengwá (K) and Mizighi (Z). One of the topics of conversation was former traditions. K:

ká

va-agirú vtu vé-tá-dára nhtúngo ni-cho ká. íji te-tú-ló ki-háli. DEM2 2-male 2-our 2-EVI-grasp gourd COP-7:REL DEM2 now NEG-I PL-have 7- other

W:

'Those men of ours hold the gourd that is thát one. Now we have nothing else.' m-kú-hé aré te-hé-ló vigí vi-hálí ní ku-vína bí. 2PL-C0ND-arrive there NEG-16-have things 8-other is 15-dance only

K:

'When you arrived there, there were no other things, it was only dancing?' kuvína mtarágha u-dárá-í-w-e. mtarágha u-dará-í-w-e ni 15-dance ñute 3-grasp-APL-PAS-PF ñute 3-grasp-APL-PAS-PF by

wá ki-agirú na unywi m-vína. they-C0N 7- male with 2PL 2PL-dance

Z:

'To dance, a flute was held. A flute was held by the men, and you dance.' va-agirú va-lagé ha-wé na kúne. hé-té-ló éhla 2-male 2-women 16-one with 2PL 16-EVI-have tradition

te-hé-ló háta mŋendo bí sa yá mgeŋenda kigí NEG-16-have even movement only as this step thing

li-vín-w-a háta mrjendo bí. 1 1-dance-PAS-F even movement only.

vé-múru ve-ku-'í ve-ku-vína hé-mínda. 2-can 2-C0ND-lie 2-COND-dance 16-house

K:

'Men, women together with you. There was traditional dance, there was a movement like that step. The traditional dance was danced with just one movement only. The dance could be danced inside the house.' hé-ku-hlé tu-ta-gále íji ní chó ká. 16-15-dawn 1PL-EVI-bring now is 7:REL DEM2

'At dawn we leave. Now this is the way it is.'

206

CHAPTER 9

Conversation on the future of Mbugu language (IMb) The following excerpt is from a conversation between Yambazi Immanuel (A) and Saidi Salimu (B) in Emao (Fwizai-Rangwi), on 16.10.1993. It was prepared and interrupted by me (I). It starts with a discussion on the prices of agricultural products. B:

A:

B:

A:

B:

A:

B:

A:

B:

A: B:

A:

B:

n-áa-ó-i-w-a mia na hamsíni. 1SG-PST-give-APL-PAS-F hundred with fifty 'I was given hunderd and fifty.' ni-té sa i'í u-kurú axoní 1SG-say like here 2SG-cultivate what 'I say what do you cultivate here?' n-kurú viazi na magalá. he árí u-bo'i nyamálo miná 1SG-cultivate potatoes with maize he 2SG 2sG-do work which 'I cultivate potatoes and maize. And you what work do you do?' án nitakurú, 1SG 1SG-EVI-cultivate 'Me, I cultivate.' u-kurú axóni 2SG-cultivate what 'What do you cultivate?' viázi, magalá. potatoes, maize 'Potatoes, maize.' magalá á ku-ho-ij-a-lé maize 6:C0N 15-go-CS-F-where 'Maize, to bring where?' á ku-'á 6:C0N 15-eat 'For consumption.' te-hé-lo á-kú-di? NEG-16-have 6:CON 15-sell 'There is none to sell?' á'a no te-hé-lo kigí ki-háli chá u-murú-kúru NEG-16-have thing 7-other 7:C0N 2sG-can-cultivate 'There is nothing else that you happen to cultivate?' tu-kurú kabichí na machagá vi-'axoní ni-vyo tu-kurú-ye IPL-cultivate cabbage with beans 8-what COP-7:REL IPL-cultivate-PF ' W e cultivate cabbages, beans, and what else? This is what we cultivate.' viazí u-dí iguniá bei mína potatoes 2sG-sell bag price which 'Potatoes you sell for how much per b a g ? '

SYNTAX, CODE-SWITCHING AND TEXTS

A:

207

élfu nú na mia sabá thousand(Sw) two with hunderd seven(Sw)

: A: B: A:

'2700' machagá 'beans' machagá 'beans' igunia 'per bag' debe ni elfu nú na mia kói kwa bei wa'íji tin(Sw) COP thousand(Sw) two with hunderd four with price now

áma ki-bakóri shílingi hamsíni or(Sw) 7-cup shiling(Sw) 50(Sw)

B:

'One tin (of beans) is two thousand five hundred at the current price, but fifty shillings for a mug.' na kwa igúnia, igúnia

A:

kwa igúnia

and with bag bag per bag

B:

ishiríngi mé shiling how:many

A:

'B: And per bag? A: Per bag. B: How many shilings?' karna debeni karna elfu nú na mia xói, xai ni sita, elfu ixóda na kói like tin-cop like thousand two with hundred four if by six thousand ten and four.

I: A:

'If a tin is 2500 (shillings), times 6, that gives (you) ten thousand and four (hundred). Kimbugu kitakuwaje? (Sw) 'What will Mbugu be like?' i-tégeméa, i-téka va-hé. hé-lo va-hé va-háli te-véyo-iy-a va-milo kíni ki-Ma'á. 9-depend 9-take 2-person 16-have 2-person 2-other NEG-2-speak-APL-F 2-child their 7-Mbugu

hé-lo vaháli hata ya-xa aré he muarjá, e-yo'-íy-a va-mílo kú'u ki-má'a ku-bá 16-have 2-other even this-16 there 16 village 1-speak-APL-F 2-child his 7-Mbugu 15-say

si rahísi ku-láhati. aré ni ... kwa wa'ijí chó tu-só hú vé-éndelea ku-zoxó. NEG:COP easy 15-get:lost. there is with now 7-REL IPL-go they 2-continuel5-stay

te-ki-ká-la bali hú ve-lí gahána kí-ne-lá NEG-7-csEC-get:lost but they 2-come:out outside 7-EXPEC-get:lost

mirá kwa sabábu te-tu-ka-té-li va-xabúna si rahisi ku-lá. but with reason NEG-lPL-NOT.YET-EVI-come:out 2-all NEG.COP easy 15-get:lost

B:

'It depends. It takes the person. There are some people who don't speak it, Mbugu, to their children. There are others here, even in this village (Emao) who speak it to their children, Mbugu, so in that case it is not easy to lose. That is the way we go; if people don't continue to stay (in this area), it (the language) will get lost. But because we don't leave all (from this area), it is not easy for the language to get lost.' áni ni-áxo ta-chí-ka-lá kwa sabábu kíra nyaxé e-sáywa i'í me 1SG-see NEG-7-NOT.YET-get:lost with reason every child l-be:born

208

CHAPTER 9 ka e-sáywa ni m-Ma'á e-ra-yo-í-w-a ki-Ma'á. here time l-be:bornCOP1-Mbugu1-BGND-speak-APL-PAS-F7-Mbugu

íi lugha ki-ka te-chí-ka-lá sari kwa madilíko kwa dunia 9:DEM1 language7-CSEC NEG-7-NOT.YET-get:lost until with changes with world

B:

'As I see it, the language will not get lost because every child that is born here is born a Mbugu. It will be spoken to in Mbugu. This language, beyond doubt, will not get lost unless the world changes.' Miaka mingapi? 'How many years?' hahali mikó ishirini i-lit-íye ki-múru ku-lá. ni ki-kuhló.

A:

'Otherwise, in twenty years the language may get lost. That is no problem.' há ki-Ma'á híle ité ni ki-kuhló ku-kí-'o sa níne va-Ma'á.

I:

16-other years twenty 4-come-PF 7-can 15-get:lost COP 7-good

DEM2 7-Mbugu or that COP 7-good 15-7-speak like we 2-Mbugu

A+B: B:

'It is good that we speak this Mbugu because we are Mbugu.' ni kikuhló. 'It is good' kwa ite níne ni lúgha kánu. ni kuhló níne ni u-yó tu-hleru-íye na-ú. with that weCOPlanguage ourCOPgood we COP 2sG-speak 1PL-meet-PF with-1

A:

'That this is our language. It is good for us that you speak, that we met.' haláfu kuhúsu ku-lá ki-murú ku-lá kwerí kwa sababu wa'iji and(Sw) about 15-get:lost 7-can 15-get:lost really(Sw) with reason now

va-he vá-chánganyika va-dóro na he va-sitané na vá-dora na hé va'ariyé, 2-person 2-mix 2-take:out with 16 Shambaa with 2-take:out with 16 2-Pare

ne vadóri vasitané, va-dóri va-roró kukuba i-tégemea iji hú va-mílo with 2-take Shambaa 2-take 2-plains that 9-depend now this 2-child

u-ne-vá-yó-iy-a hú ki-Ma'á íle u-ne-va-'o-y-a kiswahili. 2sG-EXPEC-2-speak-APL-F this7-Mbugu that(Sw) 2sG-EXPEC-2-speak-APL-F 7-swahili

u-ku-ta-va-'o-y-a ki-swahili a-ki-ta-ghá'a háli. m-lagé-ké chi u-sí mmá'a. 2sG-C0ND-EVI-2-speak-APL-F 7-swahili 1-coND-EVI-kill other. 1-wife-his 7-2SG-end 1-Mbugu

ki-Ma'á ch-áa-lá mirá kái u-ne-ra-vá-o-í-a hú ki-Ma'á. kubá 7-Mbugu 7-PST-get:lost but if 2sG-EXPEC-EVI-2-speak-APL-F this 7-Mbugu that

na hú mlagé u-ne-ra-m-xwá'-ij-a na-ú m-Ma'á. with this 1-wife 2sG-EXPEC-EVI-l-return-cs-F with-1 1-Mbugu

I:

'And about this possibility of getting lost, it is true because now-a-days people mix and marry Shambaa or Pare people. And they marry Shambaa and people of the plains. It depends now on these children; (it depends whether) you talk to them in Mbugu or Swahili. Say if you talk to them in Swahili, they will kill the language maybe. This wife ..., if you stop to be Mbugu, the language will get lost. But if you talk to them in this Mbugu language, that is, with this wife, you will turn her into a Mbugu.' Wataendelea kuzungumza kishambaa au kiswahili tu? 'Will they continue to talk Shambaa or Swahili?'

209

SYNTAX, CODE-SWITCHING AND TEXTS

A:

wataongea

kiswahili.

hata

wa'íji

ve-kumúre

te-vé'il-íye

he-lo

va-hé

they:\vill:talk swahili even now 2-many NEG-2-know-PF 16-have 2-person

ve-sé-y-w-a vaMa'á kabisa mirá te-vé-ki-'ilí-ye 2-call-APL-PAS-F 2-Mbugu completely(Sw) bul NEG-2-7-know-PF

na ve-ku-se-i-w-a va-Ma'á ihlé. with2-C0ND-call-APL-PAS-F 2-Mbugu name

ve-yo ki-sitáné kwa sababu a-xó-i-w-e twái hé va-sitáne 2-speak 7-shambaa with reason 1-go-APL-PAS-PF there 16 2-shambaa

mirá ukóo hú'a ni wa kíMa'á. but clan(Sw) his is 11 :CON 7-Mbugu

'They will talk Swahili. Even now there are many who don't know (Mbugu); there are many people who are called Mbugu but they don't know Mbugu and still they are called Mbugu by name. They speak Shambaa because they went there among the Shambaa but their clan is Mbugu.' Oral history (Normal Mbugu) The following text was recorded by my assistant, William Mboko, in Malibwi, on 17.2.92. The speaker is Mr. Ishika Mbegha. Both Ishika and William know both In­ ner and Normal Mbugu but, on my request, William posed his questions in Normal Mbugu and the answer, the monologue is in Normal Mbugu. va-mbúghu tu-ku-vóka ku-za tu-fum-íye na lukupia. 2-mbugu 1PL-COND-start 15-arrive 1 PL-come:from-PF with Lukipia

lukupia tu-ku-fuma lukupia. n-enzi yá kaé jí-lá. Lukipia lPL-COND-come:from Lukipia COP:time 9:C0N long:ago 10-DEM3

'We Mbugu, in the beginning we came from Lukipia. When we came from Lukipia, it was long ago/ tu-aa-ta-kola vaborá masa tu-ka-kiny-w-a tu-ka-díndika na 1PL-PST-EVI-meet girls problems 1PL-csEC-chase-PAS-F 1PL-CSEC-RUN with

vá-la vaborá sa chí-la kinhu tw-a-tí-w-a hé-na vighiriyá. 2- DEM3 girls as 7-DEM3 thing 1 PL-PST-say-PAS-F 16-with expulsion

'We had a court case about girls, we were chased away, and we ran away with those girls, the whole day. This is what we were told: expulsion.'

(...) tu-ka-jinhka na-yé, tu-ka-yoa Vudée. tu-ka-vata tu-ka-íngia na Vudee IPL-CSEC-ĩ with-l:REL 1 PL-csEC-go Vudee 1PL-csEC-walk 1PL-csEC-enter with Vudee

'We ran with her (these girls) and we went to Vudee. We walked and arrived in Vudee.' Vudee tu-ka-báj-ik-a. tu-ka-neta Vudée, tu-le-vóna tu-neta Vudée, Vudee lPL-cSEC-break-ST-F lPL-CSEC-be:many Vudee 1PL-BGND-see lPL-be:many Vudee.

210

CHAPTER 9

naanii (uu nee) u-ku-ra-síkia hé-ti-w-a hé-na Nhkándu, m-nhkándu na m-lóndo whatyemacallit (u ne) 2sG-C0ND-BGND-listen 16-say-PAS-Fl6-with Nhkandu 1-Nhkandu with 1-Londo

tw-ée-ré kinhtu íchí ki-mwe tu va-mbughu vóse lPL-PST-be thing 7:DEM1 7-one 1P 2-mbugu 2-all

'In Vudee we spread (in different directions). We were with many people in Vudee. When we saw that we were many in Vudee... Do you know a place called Nhkandu, Nhkandu and Londo were one, all of us Mbugu.' e-ká-ítang-w-a mwambéni m-ŋhangwaná ní-re-yé m-lóndo 1 -CSEC-call-PAS-F Mwambeni 1 -Nkangwana COP-BGND- 1 :REL Mlondo

a-za-íkaa hé lungorí lwáa wa-ŋhangwaná wakatí uo 1-IT-stay 16 hill 1 l-CON 2(Sw)-Nhangwanatime(Sw) 3:DEM2

íngeí va-lá vanhtu nhkóndo ji-ne-kúnda ku-za-ku-dósir-ij-a now 2-DEM3 persons war 10-EXPEC-want 15-IT-COND-follow-APL-CS-F

hé-ti-wa hé-na itáwí 1-aa-bájika uko Vudée la vambughu. 16-say-PAS-F 16-with branch 5-PST-divide there Vudee 5:CON 2-mbugu

'(There was a leader.) He was called Mwambeni, from the Nhangwana clan, that is from the Londo clan, and he stayed at the mountain of the Nhangwana that time; now, those people wanted war to follow them; there was a branch (of people) that divided there at Vudee, a branch of Mbugu people.' ni-hó úswi tu-ne-tónga ú-vo u-manya va-mbughu n-ée-ré-vo COP-16:REL we 1PL-EXPEC-go 3-8:REL 2G-know 2-mbugu is-PST-be-8:REL

vanangó vanangó ve-ka-bádirika va-ka-dósa sura ja vashamba. 2-Nango Nango 2-cSEC-change 2-csEC-follow face 10:CON shambaa

'So that is how we went. You know, Mbugu people includes the Nango clan. And the Nango changed and became to look like the Shambaa.' úswi tu-ka-tónga huvo nyika dú tu-ka-tónga tu-ka-fúm-i-a we 1PL-csEC-go such low:lands only 1PL-csEC-go 1PL-csEC-come:from-APL-F

Mlógolo úswi va-lóndo va-nhkándu va-ka-chwá íi na vanangó ve-ka-fúmia Lugúrwa, Morogoro we 2-Londo 2-Nhkandu 2-cSEC-cut 9:DEM1 with Nango 2-CSEC- come:from-APL-F Luguru

vé-ka-baja Shume ve-ku-baja Shume ve-ka-baja Mlolá Mbwoi 2-cSEC-split Shume 2-C0ND-split Shume 2-csEC-split Mlóla Mbwoi

'We then went to the low lands. We went on to Morogoro (=city far from Usambara). We, the Londo. The Nhkandu went past the Nango. They arrived in Luguru (=area far from Usamabara), they split in Shume (map 2). They arrived in Shume, Mlola, and Mbwai.' ve-ka-baja íli isanga úo MwendaKiró ni-ye mfúmwa hé íli isanga mpaká Bagha. 2-csEC-split 5:DEM1 land 1 :DEM2 MwendaKiro COP-l:REL king 16 5:DEM1 land until 

hambú MwendaKiró m-ku-síkia hé-na m-gonja. m-gonja ní-re-ye m-nhkandu; because MwendaKiro 2PL-C0ND-listen 16-with 1-Gonja 1-Gonja BGND-1:REL 1-Nhkandu

SYNTAX, CODE-SWITCHING AND TEXTS

211

m-gonja ni ha-la a-ra-túa Gonja ni nyumbá n-dori y-a-tua Gonja. 1-Gonja C 16-DEM3 1-BGND-live Gonja COP house 9-small 9-PST-live G

'They arrived in this land of this man called Mwenda Kiro. He was the king in this land, up to Bagha. Because Mwendo Kiro — do you listen? — he is someone from the Gonja clan. The Gonja clan is related to the Nkandu clan. The Gonja people live in Gonja, it is a small clan (house), it is in Gonja.' hé-na va-gonja va-bagha ve-ítangwa va-kirindí ní nyumbá ya ahó hé Mtóe 16-with 2-Gonja 2-Bagha 2-called 2-Kilindi is house 9:C0N here 16 Mtoe

na vá-la ve-hó Bagha va Mlwatí hé-na va-bagha vá Shume vá-ra-oká. with 2-DEM3 2-16:REL Bagha 2:CON Mlwati 16-with 2-Bagha 2:CON Shume 2-BGND-be

vá-remía hé valondo va-ka-dósa ngasu já valondo vá-shigha ji-la ja va-Nhkándu 2-stay 16 Londo 2-csEC-follow culture 10:CON Londo 2-let 10-DEM3 10:CON 2-Nkandu

hé-na mapáto va-ku-gwisha ávo va-bagha nyumbá y-akwé nambái 16-with results 2-COND-get they 2-Bagha house 9-his Nambai

'There are Gonja people (or Bagha) that are called Kilindi. That is the clan from here, at Mtoe, and those Bagha are like Mlwati. The Bagha people from Shume were Londo. They followed the culture of the Londo. They departed from the culture of the Nhkandu. That was the outcome of those Bagha: the clan of Nambai.' ávó ava va-shúme va-ku-kéta vé-ku-óka vé-na ŋombe. vé-za-óm-an-a vé-za ve-vá-itia they 2:DEM 1 2-Shume 2-C0ND-have 2-C0ND-be 2-with cow 2-IT-beat-REC-F 2-arrive 2-2-take

mapato ve-va-táha ch-a-shigha háta hé ngasu va-ghenda-zé va-va-díndika results 2-2-take 7-PST-let even 16 culture 2-do-Q 2-2-run

Those Shume, they had cows! They fought among each other, it gave them profit. They took them, it influenced even their culture. What did they do? They ran away.' ni-hó ve-za-bája bagha vé-hó ve-íkaa Vúgha ve-ku-fuma Vugha ní-hó ve-yo COP-16:REL 2-IT-split Bagha 2-16:REL 2-stay Vugha 2-coND-come:from Vugha COP-16:REL 2-go

ha nahí bagha ve-baja bagha vé-itangwa va-kirindí va bagha na vá-ngi 16 where Bagha 2-split Bagha 2-called 2-Kilindi 2 Bagha with 2-other

vé-ítangwa va-Fúti nyumbá y-akwé Nambái. ni vanhkándu ávo. 2-called 2-Futi house 9-his Nambai COP 2-Nkandu they

úswi tu-ku-fúma kúlá tu-ku-fúma Zighula we lPL-COND-come:from there 1PL-COND-come:from Z

'The Bagha split up. Some stayed in Vugha. When they left Vugha, they went where? To Bagha. They passed Bagha; they were called the Kilindi of Bagha. And others were called the Futi people, the clan of Nambai.They were Nkandu. When we came from there, we came from Zigua.'

212

CHAPTER 9

Oral History (Inner Mbugu) The following excerpt is from the history of the Mbugu according to Mr. Sewia, recorded in Februari 1993 in Magamba; text in Inner Mbugu. tú-kú-lí na luhíza tu-kú-líta tu-fíkire vudé. ŋ-hó ve-líta vamasái, 1PL-coND-originate with luhíza 1PL-coND-come 1PL-arrive vudé COP-16:REL 2-come Masai

tw-áa-zóko múda mw-ezá, ve-líta vamasái. va-ku-líta tó vé-dóri wá na afá na ma'álú 1PL-PST-sit time 14-long 2-come Masai 2-COND-come there 2-take:out cattle with goat(s) with sheep

vé-á-teka hú vaMa'á vé-bá: á tu-ké-we ní ahoní i'í. tú-dóse wá kánu. 2-6-take DEM2 Mbugu 2-say aa 1PL-put-PAS:PF by what here lPL:D-follow-SBJ cattle our

vé-dósa wá kíni na hé-vamasái. vá-kú-hé vé-ómba kázi. vé-dúmu kázi 2-follow cattle their with 16- Masai 2-COND-arrive 2-request work 2-want work

tú-mú-zéú-i-yé wá. vé-bá há tú-mú-'ó kázi nne. vé-zá-zéu v-áa-zóko ma-'azé má-ngi. lP:D-2PL-herd-APL-PF cattle 2-say haa lP:D-2PL-give day 4 2-IT- herd 2-PST-sit days 6-many

mirá íjí vé-dúmu akíri ya ku-dór-íy-a wá. basi vé-dóri vé-bá-w-e but now 2-want brains 9:CON 15-take:out-APL-F cattle well 2-take:out 2-say-PAS:PF

vá-géra wá sa ikúmi vé-bá-w-e m-ka-na-chéleja wá tó sari sa ikúmi ni ahoní 2:D-bring cattle hour ten 2-say-PAS-PF 2PL-CSEC- NOT.YET-late cattle there until hour ten COP what

wá tu-zéu-y-a úda. ahahah kái m-zéu-y-a úda mirá mú-si-chélíj-é cattle 1PL-herd-APL-F far aha if 2PL-herd-APL-F far but 2PL-NEG:SBJ-late-NEG

wá tó. tó hlenú vé-jí-géra sa sita, vé-bá aké vé-bá-w-e cattle there there tomorrow 2-10-bring hour six 2-say because 2-say-PAS-PF

aaa bási il-éni m-zéu. bási vé-té-za-zéu sari tó hlenú. aaa well know-PL.ADR 2PL-herd well 2-EVI-IT-herd until there tomorrow

haya vé-galé-ti chá magherí sa nú, sa kái, sa hái, sa kói, sa sita, sa saba, haya 2-bring-cs at evening hour two hour four hour four hour five hour 6 hour 7

sa mnane, sa kendá, sa ikúmi aaa v-áa-pata venyéji vá ki-zughule hour eight hour nine hour ten aaa 2-PST-get inhabitants 2:C0N 7-Zigua

mbánye m-tú-ghenj-é ku-tahónhi wá kánu i'í hé-ruvú. well 2PL-lPL-help-SBJ 15-pass:cs cattle our here 16-Ruvu

'We came from Luhiza and we arrived in Vudee. That is when the Maasai came; we had stayed a long time when the Maasai came. They came there, they took our cattle, goats and sheep. The Mbugu said: "What are we waiting for here? Let us follow our cattle." They followed their cattle to the Maasai. When they arrived, they asked for work. They wanted work. "Let us herd cattle for you." They said: "Haa, we should give you work, we?!" They went to herd the cattle and they stayed (among the Maasai) for a long time. But then they were looking for a trick to steal the cattle back. When they took the cattle out (to herd), they were told to return the cattle at four

SYNTAX, CODE-SWITCHING AND TEXTS

213

o'clock. Then they were told: "Why do you make it so late? The cattle is to be back at four o'clock?" "We herd the cattle far away." "Ahahah it is ok if you herd far away. But don't let the cattle be late." The next day they bring the cattle back at twelve o'clock. They said (because they were told): "You should know how to herd." Well they went to herd there the next day. When they were supposed to return the cattle in the evening, it became eight o'clock, nine o'clock, ten o'clock, eleven o'clock, twelve o'clock, one o'clock, two o'clock, three o'clock, four o'clock. (They did not come back with the cattle.) They got local people, the Zigua, (to help them). "You help us to let the cattle pass through the Ruvu river."'

NOTES

Notes to Chapter 1 1 Both Bastin in her Bantu bibliography (Bastin 1975) and Whiteley in his 1960 article make this mistake. 2 "hier lag die seltsame Tatsache vor, dass die eigentlichen Wortstämme offenbar nicht Bantucharakter tragen, dass aber trotzdem die Bantupräfíxe vor diesen Wortstämmen in vielen Fällen erscheinen" (Meinhof 1906: 294), and "Der Bantuisierungsprozess, in dem die Sprache sich befindet, ist sehr charakteristisch und zeigt an einem Beispiel, wie eine Sprache, die ur­ sprünglich zu ganz anderem Stamm gehört, durch Umflutung mit einer fremden Grammatik allmählich umgebildet werden kann." (Meinhof 1906: 296). 3 The title mentions the Kibwyo dialect while the relevant chapter identifies the language as Mbugu. 4 Note that this example involves reinterpretation of the initial syllable in the Afrikaans source as a prefix for plural class 6 and backformation of the singular class 5 prefix le5 Pa stands for Pare, Sh for Shambaa. 6 Metathesis of syllables may be the origin of Sagala hasi 'tendon' from *sipa. 7 Ehret (1980) does not distinguish Mbugu and Ma'a roots for his reconstructions. Notes to Chapter 2 1 Those for which no source is given but only a place name and a date were collected by my­ self. 2 The "second sea" most propably refers to the Maasai planes of Tanzania. The savannah planes are referred to as "sea". 3 I encountered a remarkable instance of this phenomenon when I wrote down the oral history of the Alagwa (a Cushitic group in Tanzania) from one of the most respected elders. His first sentence mentioned that they are from Maka (=Mecca), which reflects the fact that the Alagwa are Muslims; they became Muslim at the beginning of the 20th century following their chief. In his next sentence he jumped to the present-day Alagwa area in Northern Tanzania. 4 The evidence that is presented is mainly from Iraqw because there are very few data on Gorwaa. 5 The verb 'to hunt' in Shambaa is ku-ingafa. 6 The Zigua data are from my own field notes. 7 They call themselves Bowe.

216

NOTES

8

There are soda lakes in the area where salt can be obtained in the dry period. The words Taita and Davida [Daßida] actually go back to the same root. 10 According to Philippson (p.c.) this is a lateral realisation of r. 11 Sagala has also kirawa 'arrowhead'. 12 Cf. Davida zoya 'to begin'. 13 This is a problematic regional word (also in Chaga), cf. Pare baa 'to break' but the tone indi­ cates it is a loan in Pare. 14 In Sagala now sarna, from Mijikenda. Wray (1894) also has tsarusa ngolo 'to comfort'. 15 However, IMb shighatí may also be related to IMb shigha 'leave behind' which is Pare and also Sagala siga, Davida shigadi 'outside' (Tate 1904:145). 16 This must be a loan in Sagala because of the initial t. 17 Sako means 'blood' in Inner Mbugu but it is not evident that this is the same etymon. 18 Walsh (1990:76-77) provides further evidence for an association between the Degere and the Taita Cushites. 19 Meinhof (1906) has suggested that this word is derived from ahla 'fire' plus a denominalising suffix -ha, i.e. the standard Bantu denominalising suffix -pa. 20 Another plausible option is that it is derived from the Inner Mbugu word ma 'áw 'secret'. 21 Gerda Rossel (p.c.). 22 In Pare the word refers to any plantation; in Shambaa specifically to a banana plantation. 23 It is common that the words for ensete and for bigger seeds such as maize and sorghum are identical (Gerda Rossel, p.c.). 24 Dorobo is a cover term for all dependent groups under the Maasai irrespective of their eth­ nicity. They have no cattle and maintain a fixed and unfree trading relation with the Maasai. 25 The correspondence gu - gwe is common in Pare.

9

Notes to Chapter 3 1 These recordings are kept in the Berliner Phonogramm-Archiv of the Ethnologisches Mu­ seum in Berlin-Dahlem and I listened to them. Unfortunately, the quality of the Mbugu re­ cording is too poor to recognise anything but noise. 2 The sound is also reconstructed for Southern Nilotic (Rottland 1982). 3 The loss of the lateral fricative in the non-Southern Cushitic languages is taken to be an inno­ vation that defines these languages as separate from Southern Cushitic (Ehret 1980, but see Hetzron 1980). 4 Green also has le in the first item but this form is, as he pointed out, from Shambaa le 'long'. The interpretation of Storch's data is not without problems, because it would entail that he uses ss for while using sh, s, shl for a further problem is t for in Fader's data. 5 The juxtaposition is supposedly just like Iraqw and Burunge. In fact Burunge needs a linker word and Iraqw has juxtaposition but the first noun, the possessed noun, needs to be marked for the status constructus involving a gender linker.

NOTES

217

6 Before him Farler noted milo for mmilo, and halu, pl: mi- for Valu 'sheep' — in general he has h for '—, however, he also has muhafa 'goat', which is now afa and maratha 'river', which is now haráza. 7 The Bantu language Gweno, spoken in the Pare area, has a verb 'to have', -úra. 8 With the exception of káti 'kindle'. 9 Some exceptions are 'aghati 'compare', firiti 'finish', darati hold tied, tati 'cheat'. 10 Exceptions are gáléti 'bring', hokwáti 'be late', kwetí 'quarrel'. 11 Cf. Iraqw and West-Rift Southern Cushitic ki?. 12 Other Inner Mbugu causative derived verbs may be hlúku 'drop' from hlú 'fall', bútu 'put into' from búu 'enter', hlutú 'pass on, let inherit'. 13 One of those is the extremely rare: *-Vx (Ehret 1980:64,107), which occurs as -Vk in Dahalo but only in combination with a following stative, continuative or causative extension, see Tosco (1991:46). Outside Dahalo it only occurs in Ma'á: lud-uxu 'to lean', which I don't have in my data unfortunately, and in the noun kwacaho [= nhkwachoho] 'calabash cleaning stones', but this word is not Inner Mbugu. 14 With the exception of the word kisimbáno (I+NMb) 'summit', from Shambaa kisimba 'lid of a small basket', such endings are limited to IMb. 15 Ehret (1980:53) reconstructs Proto Southern Cushitic -eta on the basis of Burunge, Aasáx and Ma'á and a suffix -eto on the basis of Iraqw and Kw'adza -eto, Asa -et, and Dahalo -etto. However, there is no Iraqw suffix -eto. 16 The NMb nouns idórome 'ram', sakame 'blood' (cf. sako (IMb)), and the IMb noun kusáme 'leg' and the adjective kuhamé 'black' do not contain this suffix. 17 Tucker and Bryan also have dilaú for mdiláo 'king' (IMb). 18 The ending is not to be confused with a Bantu nominalisation in -o as in mgwalo 'thief (IMb) from gwa 'steal' retaining an intervocalic l. 19 Goats are kept inside the house where the fire is. 20 But ka for feminine nouns and not ti. 21 G stands for gemination of the following consonant. 22 The Bumbuli form for the second person, kechi, was also reported for Ma'á by Green (1958). 23 She also mentioned in private communication that there is the possibility that the situation at the time of my fíeldwork in the 1990s was radically different from that earlier in the 20th cen­ tury. I don't think this is the case. I find no evidence for any dramatic change during the docu­ mented history of Ma'á; on the contrary, I see a remarkable stability. The differences between my description and those by earlier researchers are due to their failure to pay attention to the non-mixed language of the community and therefore to its relationship with the mixed variety, despite the fact that some of them mention its existence (see section 3.2). 24 I assume that all Cushitic m/f/pl agreement was gradually lost in the period when command of the pre-Ma'á language was diminishing.

218

NOTES

25

Dahalo distinguishes between masculine and feminine in second and third person singular possessives, Boraana only in third person singular, and the Iraqw group languages do not dis­ tinguish sex in possessives at all. 26 The number distinction is not always present in the non-human category. 27 In addition, Dahalo and some other Cushitic languages have subject and other inflection in a pre-verbal element (termed selector in Dahalo and the Iraqw-group and indicator particle or subject clitic in Somali). Similar particles exist in the Konso and the Dullay groups within Eastern Cushitic. 28 I cannot offer an explanation why the Cushitic passive was not retained while the causative was. A complicating factor is that the developments of the Cushitic passive in Dahalo and in the Iraqw cluster are drastic and unclear. 29 See Nurse (2000) for a similar argument for the reverse influence, the development of a pro­ liferation of preverbal  indicators in the Bantu language Rimi under influence from the Iraqw-group. 30 Note, however, that that is precisely what often happens in Swahili - English as well as in other forms of code-switching (see Myers-Sotton ( 1993a:110ff, 189f)). 31 Ehret considers it productive but I don't, see Thomason (1997:481, note 8) 32 "... das Mbugu seiner genetischen Stellung nach in der Tat als eine Mischsprache zu betrach­ ten ist. Das Mbugu ist jedoch nicht die einzige Sprache, die in diesem Sinne als Mischsprache angesprochen werden kann. In der gleichen Weise ist z.B. die armenische Zigeunersprache nach der Beschreibung von Meillet als Mischsprache zu bezeichnen (Meillet 1914: 420)" (Heine 1969:1111). 33 For this same reason I do not see the need for providing a more precise definition of "mixed language". 34 In the case of Mbugu, such a development of the loss of the lexicon of the language donating the grammar is conceivable since some of the communities hardly ever use Normal Mbugu. 35 In my view, the coexistence of Ma'á and its matrix language Normal Mbugu is difficult to imagine. Ma'á is the result of a shift that does not reach completion, but at the same time peo­ ple do speak Normal Mbugu and thus the shift did go to completion. Notes to Chapter 4 1 Two exceptions are nj in verb forms such as émanje 'he knew' (NMb), based on the root many, that have an inflectional final vowel e, see 5.1, and the voiceless palatal pre-nasalised stop in the word nyh(c)upa 'bottle'. 2 Leaving aside the case of the Normal Mbugu word for 'no' which is a'a and does contain a glottal stop. 3 There is one counter-example, ijengho (NMb) 'tooth'. 4 This is according to Philippson's data. The earlier source, on Pare, Kotz (1909), has only g and not gh but this might also be a dialectal difference. 5 Storch (1895) had viyauhu.

NOTES

219

6

Storch (1895) had the intermediate form with g, muharega. With the verb vwnbn 'to rot' these are the only recorded examples of the sequence consonantglide-round vowel. 8 Nouns in e.g. class 9 and 10 are excluded in these observations since for these nouns the nominal prefix and root-initial consonant are fused. 9 Note that the numbers do not completely add up because there are some nouns for which I do not know to which of the two languages they belong. 7

Notes to Chapter 5 1 "Tense" is used here and elsewhere in this chapter in its traditional sense, i.e. to mean a tenseaspect-mood-focus expressing inflection unit of the verb. 2 Note that not all the examples from Normal Mbugu that are taken from texts are tone marked. 3 With the exception of inhka 'give' the verbs with initial i cannot be considered to be histori­ cally derived from a CVC root. 4 This is a reflex of Common Bantu yíb, Proto-Sabaki iW (Nurse and Hinnebush 1993:550). 3 It is not used in the expectational. which might be considered non-assertive based on its se­ mantics. 6 In imperatives sa is used as an itive marker. 7 There is, however, at least one defective and auxiliary verb from Arabic origin in Swahili, tasa 'not yet', whose final vowel is always a, see Sacleux (1939) and Nurse and Hinnebush (1993:556). 8 In addition, IMb has verbs that end in a vowel other than a, that is dropped before the perfect ending, e.g. -zíe 'you have come' from zó (IMb),  'it knows' from 'ile (IMb). 9 But manga SPEED in the same position is never menge. 10 This morpheme represents an innovation among the Bantu languages of the region and is shared only by Pare and Taveta; a problematic possible cognate nta occurs in some Swahili dialects. If this turns out to be correct, the negative prefix was at some stage in history not ex­ cluded from this a/e alternation, see Nurse & Hinnebush (1995: 469-470). 11 Cf. the IMb question word kimómo 'how'. 12 There are exceptions among passive perfect verb forms in e with a prefix åa. 13 The tonal spread onto the '"tense'" prefix in this example does not occur otherwise with this "tense", but is obligatory with certain other "tenses". 14 There is, however, one instance of za-ku in the texts, namely in example (77) above. 15 Both were translated in a variety of Swahili forms: nitakapolima, nikijalima, nikilima, or nilikuwa nalima. This last translation equivalent was also used for n-ée-te-re-íma. 16 The suffix -é has a variant -áe attested twice only: u-si-iv-áe (NMb) 2SG-NEG-forget-NEG 'don't forget', t-é-ka-ghat-áe (NMb) NEG-1-NEG.FUT-tired-NEG 'he won't get tired'. 17 There is another verb 'oshidnwa (IMb), which also means 'to dream' and contains a passive derivational suffix as well.

220

NOTES

18 The Inner Mbugu non-productive causatives -ti and -ri are an exception in this respect since they end in a vowel. 19 This shows that the long vowel consists of a sequence of a root vowel and an inflectional vowel. 20 This form contains the reflexive object prefix ku, see section 5.1.2.

Notes to Chapter 6 1 I have only these two examples of 11/10 words with a stem-initial s in my materials. A pair with a stem-initial sh does show a nasal in the class 10 plural form: ¡nshwizo, nshwizo (11) 'sharpened wood e.g. to pierce ear lob' (I+NMb). 2 There is in fact an archaic IMb word snghano 'right' that is not in class 15. 3 Also partitives, as in kazwari 'piece of cloth' (NMb), kavungo 'piece of cloth' (NMb). 4 The noun ahoní 'what' (IMb) can occur with a class 7 prefix. Notes to Chapter 7 1 The Inner Mbugu equivalent kuhamé is not an adjective according to my definition because it does not show agreement, e.g., mu'á knhamé 'a black head' (IMb), in which the head noun is in class 3 and the adjective has no agreeing prefix. It behaves like the higher numbers (see section 8.4). 2 This is the modifying question word 'which', not a relative pronoun. 3 The expected form is ló. 4 In fact, in my data I do have the phrase mlola yose 'all of the town of Mlola' that shows class 9 agreement on the quantifier ose. Notes to Chapter 9 1 Note that the applied extension introduced the beneficient "us" as object which in turn has become the subject of the passive derived verb leaving the original object of the verb-withoutapplied, "bull", as a "chômeur" or demoted object. 2 voka (NMb) 'begin, start' is very often used in stretches of IMb although there is an IMb counterpart which is not very current.

REFERENCES Agius, Dionisius A. 1996. Siculo-Arabic. (Library of Arabic Linguistics, 12.) London: Kegan Paul. Agostini, Francesco, Annarita Puglielli and Ciise Moxamed Siyaad 1985. Dizionario Somalo - Italiano. Rome: Gangemi Editore. Ashton, E.O. 1944. Swahili grammar (including intonation). London: Longman. Bakker, Peter. 1997. "A language of our own": The genesis of Michif the mixed CreeFrench language of the Canadian Métis. (Oxford Studies in Anthropological Linguistics, 10.) New York: Oxford University Press. Bakker, Peter and Maarten Mous 1994. Introduction. In Bakker and Mous (eds.) Mixed languages: 15 case studies in language intertwining, pp. 1-11, Amsterdam: IFOTT. Bakker, Peter and Maarten Mous (eds.) 1994. Mixed languages: 15 case studies in language intertwining. (Studies in Language and Language Use, 13.) Amsterdam: IFOTT. Banti, Giorgio 1997. Low-caste jargons in the Somali- and Oromo-speaking areas of the Horn. Paper presented at the Symposium on Endangered Languages in Africa, Leipzig. Bastin, Yvonne 1975. Bibliographie bantoue sélective. (Archives d'Anthropologie, 24.) Tervuren: Musée Royal de l'Afrique Centrale. Bastin, Yvonne 1983. La finale -ide et l'imbrication en bantoue. Tervuren: Musée Royal de l'Afrique Centrale. Bastin, Yvonne 1994. Reconstruction formelle et sémantique de la dénomination de quelques mammifères en bantou. Afrikanistische Arbeitspapiere 38:5-132. Baumann, Oscar 1890. In Deutsch-Ostafrika während des Aufstandes. Vienna: Eduard Hölzel. Baumann, Oscar 1891. Usambara und seine Nachbargebiete. Berlin: Dietrich Reimer. Besha, Ruth M. 1974. A socio-linguistic description of Kimaa and its points of contact with Kishambala. Dar es Salaam: MA thesis. Besha, Ruth M. 1989. A study of tense and aspect in Shambala. (Language and Dialect Studies in East Africa, 10.) Berlin: Dietrich Reimer. Besha, Ruth M. and Derek Nurse 1979. Shambaa. In Derek Nurse (ed.) Description of sample Bantu languages of Tanzania. = African Languages/Langues africaines 5.1, pp. 90-103. London: International African Institute

222

REFERENCES

Biber, Douglas 1994. An analytical framework for register studies. In Douglas Biber and Edward Finegan (eds.) Sociolinguistic perspectives on register, pp. 31-56. Ox­ ford: Oxford University Press. Boretzky, Norbert 1985. Sind Zigeunersprachen Kreols? In Norbert Boretzky, Werner Enninger and Thomas Stolz (eds.) Akten des 1. Essener Kolloquiums über Kreol­ sprachen und Sprachkontakte, pp. 43-70. Bochum: Brockmeyer. Brenzinger, Matthias 1987. Die sprachliche und kulturelle Stellung der Mbugu (Ma'a). M.A. thesis, University of Cologne. Brenzinger, Matthias (ed.) 1992. Language death: Factual and theoretical explora­ tions with special reference to East Africa. Berlin: Mouton-De Gruyter. Brenzinger, Matthias 1992. Lexical retention in language shift: Yaaku/MukogodoMaasai and Elmolo/Elmolo-Samburu. In Matthias Brenzinger (ed.), Language death: Factual and theoretical explorations with special reference to East Africa, pp. 213-254. Berlin: Mouton-De Gruyter. Conte, Christopher Allan 1994. Transformations along the gradient: Ecological change in the mountains and plains of northeastern Tanzania's West Usambara Mountains,  1860-1970. PhD dissertation, Michigan State University. Copland, B.D. 1933-34. A note on the origin of the Mbugu with a text. Zeitschrift für Eingeborenen-Sprachen 24:241-245. Coupez, André, Yvonne Bastin, E. Mumba. 1998. Bantu Lexical Reconstructions 2. Tervuren: Musée Royal de l'Afrique Centrale. Cronk, Lee 1989. From hunters to herders: Subsistence change as a reproductive strat­ egy among the Mukogodo. Current Anthropology 30:224-234. Declerck, Renaat 1988. Studies on copular sentences, clefts and pseudo-clefts. (Symbolae Facultatis Litterarum et Philosophiae Lovaniensis, Series  Linguistics, 5.) Dordrecht: Foris Publications. De Rooij, Vincent A. 1996. Cohesion through contrast: Discourse structure in Shaba Swahili / French conversations. (Studies in Language and Language Use, 24.) Am­ sterdam: IFOTT. Dixon, R.M.W. 1997. The rise and fall of languages. Cambridge: Cambridge Univer­ sity Press. Ehret, Christopher 1974. Ethiopians and East Africans: The problem of contacts. Nai­ robi: East African Publishing House. Ehret, Christopher 1980. The historical reconstruction of Southern Cushitic phonology and vocabulary. (Kölner Beiträge zur Afrikanistik, 5.) Berlin: Dietrich Reimer. Ehret, Christopher and Derek Nurse 1981. The Taita Cushites. Sprache und Geschichte in Afrika (SUGIA) 3:125-168. Elderkin, E. Derek 1988. Person and number markers in Iraqw verbs. Afrikanistische Arbeitspapiere 14:79-96. Farler. 1885. Appendix to A.D. Shaw A pocket vocabulary of the Ki-Swahili, KiNyika, Ki-Taita and Ki-Kamba languages. London: S.P.C.K. Feierman, Steven 1974. The Shambaa Kingdom: A history. Wisconsin: The University of Wisconsin Press.

REFERENCES

223

Ferguson, Charles A. 1994. Dialect, register, and genre: Working assumptions about conventionalization. In Douglas Biber and Edward Finegan (eds.) Sociolinguistic perspectives on register, pp. 15-30. Oxford: Oxford University Press. Gamta, Tilahun 1989. Oromo-English dictionary. Addis Ababa: Addis Ababa Univer­ sity. Giacalone Ramat, Anna 1995. Code-switching in the context of dialect / standard lan­ guage relations. In Lesley Milroy and Pieter Muysken (eds.) One speaker, two lan­ guages: Cross-disciplinary perspectives on code-switching, pp. 45-67. Cambridge: Cambridge University Press. Goodman, Morris 1971. The strange case of Mbugu. In Dell Hymes (ed.) Pidginization and creolization of languages, pp. 243-254. Cambridge: Cambridge University Press. Gragg, Gene B. (ed.) 1982. Or dictionary. (Monograph, 12.) East Lansing (MI): African Studies Center, Michigan State University. Green, E.C. 1958. The Wambugu of Usambara (with notes on Kimbugu). Tanganyika Notes and Records 2:175-1 89. Greenberg, Joseph H. 1963. The Mogogodo, a forgotten Cushitic people. Journal of Af­ rican Languages 2( 1 ):29-43. Greenberg, Joseph H. 1966. The languages of Africa. Bloomington: Indiana University. Greenberg, Joseph H. 1999. Are there mixed languages? In L. Fleishman e.a. (eds) Es­ says in poetics, literary history and linguistics presented to Viacheslav Vsevolodovich Ivanov on the occasion of his seventieth birthday, pp. 626-633. Moscow: OGI. Guy, Gregory R. 1990. The sociolinguistic types of language change. Diachronica 7:4761. Hailey, William Malcolm 1950. Native Administration in the British African territories. Part I East Africa: Uganda, Kenya, Tanganyika. London: His Majesty's Stationary Office. Harris, Grace Gredys 1978. Casting out anger: Religion among the Taita of Kenya. (Cambridge Studies in Social Anthropology, 21.) Cambridge: Cambridge Univer­ sity Press. Hayward, Dick [Richard] 1984. The Arbore language: A first investigation. (Cushitic Language Studies, 2.) Hamburg: Helmut Buske. Heine, Bernd 1969. Zur Frage der Sprachmischung in Afrika. In Wolfgang Voigt (ed.) 17. Deutscher Orientalistentag von 21. bis 27. Juli 1968 in Würzburg, pp. 11041112. (Zeitschrift der Deutschen Morgenländischen Gesellschaft, supplementa 1.) Wiesbaden: Franz Steiner. Heine, Bernd 1975. Notes on the Yaaku language (Kenya). Afrika und Übersee 58:2761, 119-139. Heine, Bernd, Franz Rottland, and Rainer Vossen 1979. Proto-Baz: Some aspects of early Nilotic-Cushitic Contacts. Sprache und Geschichte in Afrika (SUGIA) 1: 75-91. Hetzron, Robert 1980. The limits of Cushitic. Sprache und Geschichte in Afrika (SUGIA) 2:7-125.

224

REFERENCES

Hetzron, Robert and Endre P. Talós 1982. Christopher Ehret's 'The historical recon­ struction of Southern Cushitic phonology and vocabulary': A review article. Sprache and Geschichte in Afrika (SUGIA) 4:239-250. Hock, Hans Henrich 1991". Principles of historical linguistics. Berlin: Mouton-De Gruyter. Hohenberger, J. 1929/30. Zur Lautlehre und Grammatik des Asu. Zeitschrift für Eingeborenen Sprachen 20:175-208. Huffman, Marie K. and Thomas J. Hinnebush 1998. The phonetic nature of "voice­ less'' nasals in Pokomo: Implications for sound change. Journal of African Lan­ guages and Linguistics 19: 1-19. Huijzendveld, Frans 1997. 'Die ostafrikanische Schweiz' Plantages, planters en platte­ landsontwikkeling in West-Usambara, Oost-Afrika, ca. 1870-1930. PhD disserta­ tion Vrije Universiteit Amsterdam. Johnston, Harry Hamilton 1886. The Kilima-njaro expedition: A record of scientific exploration in Eastern Equatorial Africa. London: Kegan Paul, Trench and c. Kagaya, Ryohei 1989. A classified vocabulary of the Pare language. (Bantu Vocabulary Series, 6.) Tokyo: ILCAA. Kähler-Meyer, Emmi 1962-63. Studien zur tonalen Struktur der Bantusprachen. Afrika und Übersee 46:1-42, 250-295. Kähler-Meyer, Emmi 1964. Töne und Akzente in der Formenlehre des Chasu (Tangany­ ika). Afrika und Übersee 47:89-133. Kassam, Aneesa 1986. The Gabbra pastoralist / Waata hunter-gatherer symbiosis: A symbolic interpretation. Sprache und Geschichte in Afrika (SUGIA) 7:189-204. Kießling, Roland 1998. Reconstructing the sociohistorical background of the Iraqw lan­ guage. Afrika und Übersee 81:167-225. Kießling, Roland and Maarten Mous To appear. The lexical reconstruction of West Rift (Southern Cushitic). Cologne: Rüdiger Köppe. Kimambo, Isaria N. 1969. A political history of the Pare of Tanzania (1500-1900). Nai­ robi: East African Publishing House. Kolbusa, Stefanie [1995]. Schwiegermeidung bei den Nyakyusa. Ms. Universität Bayreuth. Kotz, Ernst 1909. Grammatik des Chasu in Deutsch-Ostafrika (Pare-Gebirge). (Archiv für das Studium deutscher Kolonialsprachen, 10.) Berlin: Georg Reimer. Krapf, Johann Ludwig 1858. Reisen in Ostafrika ausgeführt in den Jahren 1837-1855 von Johann Ludwig Krapf Stuttgart: In Commision bei W. Stroh. Künzel, Wieland, Alexander Kagi, and Michael Mdoe 1982. Implications of traditional cattle-keeping: A case study of social aspects of the traditional livestock economy in two villages in the Western Usambara Mountains / Tanzania. Lushoto (Tanzania): TIRDEP. Kunene, Daniel P. 1958. Notes on Hlonepha among the Southern Sotho. African Stud­ ies 17:159-182.

REFERENCES

225

LangHeinrich, F. 1921. Schambala-Wörterbuch. (Abhandlungen des Hamburgischen Kolonial instituts, 43; Reihe : Völkerkunde, Kulturgeschichte und Sprachen, 23.) Hamburg: L. Friederichsen. Legère, Karsten 1988. Bantu and Southern Cushitic. The impact of Kiswahili on Iraqw. Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 41:640-647. Liszka, Stanley W. Jr. 1974. A preliminary report of research on the origins and internal migrations of the Taita people. Mila: A biannual newsletter of cultural research 4:38-64. Nairobi: Institute of African Studies, University of Nairobi. Maddieson, Ian 1991. Articulatory phonology and Sukuma 'aspirated nasals'. In Kath­ leen Hubbard (ed.) Proceedings of the 17th annual meeting of the Berkeley Linguistic Society, Special session on African language structures, pp. 145-54. Berkeley: Berkeley Linguistic Society. Matras, Yaron 2000. Mixed languages: A functional-communicative approach. Bilingualism: Language & Cognition 3:79-99. Mazrui, Alamin M. 1995. Slang and codeswitching: The case of Sheng in Kenya. Afrikanistische Arbeitspapiere 42:168-179. Meeussen, A.E. 1980. Bantu lexical reconstructions. (Archief voor Antropologie, 27.) Tervuren: Koniklijk Museum voor Midden-Afrika. Motingea Mangulu, André 1996. Le Lingala du pool Malebo: Éléments de phonologie et de morphologie. Afrikanistische Arbeitspapiere 46:55-117; 48:5-54. Meinhof, Carl 1906. Linguistische Studien in Ostafrika, X: Mbugu. Mitteilungen des Seminars für orientalische Sprachen 9:294-323. Merrit, E. Hollis 1975. A history of the Taita of Kenya to 1900. PhD dissertation Indiana University. Meyer, Hans 1890. Ostafrikanische Gletscherfahrten: Forschungsreisen im Kilimandscharo-Gebiet. Leipzig: Verlag von Duncker & Humblot. Mol, Frans. 1972. Maa: A dictionary of the Maasai language and folklore; EnglishMaasai. Nairobi: Marketing and Publishing. Mol, Frans. 1996. Maasai language and culture dictionary. Lemek (Kenya): Maasai Centre. Mous, Maarten. 1993. A grammar of Iraqw. (Kuschitische Sprachstudien, 9.) Ham­ burg: Helmut Buske. Mous, Maarten. 1994. Ma'a or Mbugu. In Bakker and Mous (eds.) Mixed languages: 15 case studies in language intertwining, pp. 175-200. Amsterdam: IFOTT. Mous, Maarten. 1996. Was there ever a Southern Cushitic language pre-Ma'a? In  Griefenow-Mewis and R.M. Voigt (eds.) Cushitic and Omotic Languages; Pro­ ceedings of the Third International Symposium, pp. 201-211. Cologne: Rüdiger Köppe. Mous, Maarten 1997. The e/a alternation in Mbugu: The limits of allomorphy. In Jane Coerts and Helen de Hoop (eds) Linguistics in the Netherlands 1997, pp. 123-134. Amsterdam: John Benjamins.

226

REFERENCES

Mous, Maarten 2001a. Ma'a as an ethnoregister of Mbugu. In Derek Nurse (ed). His­ torical Language Contact in Afrika. = Sprache und Geschichte in Afrika (SUGLA) 16/17, pp. 293-320.Cologne: Rüdiger Köppe. Mous, Maarten 2001b. Paralexification in language intertwining. In Norval Smith and Tonjes Veenstra (eds.) Creolization and Contact (Creole Language Library, 23.), pp. 113-123. Amsterdam: John Benjamins. Mous, Maarten To appear. The linguistic properties of lexical manipulation and its relevance for Ma'á and for mixed languages in general. In Peter Bakker and Yaron Matra (eds.) xxx Mous, Maarten, Martha Qorro and Roland Kießling 2002. An Iraqw - English Dic­ tionary. Cologne: Rüdiger Köppe. Mreta, Abel Yamwaka 1998. An analysis of tense and aspect in Chasu: Their form and meaning in the affirmative constructions. (Beiträge zur Afrikanistik, 10.) Hamburg: Lit. Mreta, Abel Yamwaka 2000. The nature and effects of Chasu-Kigweno contact. In Kulikoyela Kahigi, Yared Kihore and Maarten Mous (eds) Lugha za Tanzania / Languages of Tanzania: Studies dedicated to the memory of Prof. Clement Maganga, pp. 177-189. Leiden: CNWS. Muysken, Pieter 1997. Code-switching processes: Alternation, insertion, congruent lexicalization. In Martin Pütz (ed.) Language choices: Conditions, constraints, and consequences, pp.361-380. Amsterdam: John Benjamins. Myers-Scotton, Carol 1993a. Duelling languages: Grammatical structure in codeswitching. Oxford: Oxford University Press. Myers-Scotton, Carol 1993b. Social motivations for codeswitching: Evidence from Africa. Oxford: Oxford University Press. Myers-Scotton, Carol 1998a. A way to a dusty death: the Matrix Language turnover hypothesis. In Lenore A. Grenoble and Lindsay J. Whaley (eds.) Endangered lan­ guages: Language loss and community response, pp. 289-327. Cambridge: Cam­ bridge University Press. Myers-Scotton, Carol 1998b. A theoretical introduction to the Markedness Model. In Carol Myers-Scotton (ed.) Codes and consequences: Choosing linguistic varieties. Oxford: Oxford University Press. Myers-Scotton, Carol 2002. Contact linguistics; bilingual encounters and grammati­ cal outcomes. Oxford: Oxford University Press. Myers-Scotton, Carol and Janice L. Jake 2000. Four types of morpheme: evidence from codeswitching, Broca's aphasia, and second language acquisition. Linguistics 38:1053-1100. New, Charles 1873, 1971.3 Life, wanderings, and labours in Eastern Africa, with an account of the first successful ascent of the equatorial snow mountain Kilima Njaro and remarks upon East African slavery. (Missionary Researches and Travels, 16.) London: Frank Cass. Niceforo, Alfredo 1912. Le génie de l Argot : Essai sur les langages spéciaux, les ar­ gots et les parlers magiques. Paris: Mercure de France.

REFERENCES

227

Nurse, Derek 1979. Classification of the Chaga dialects: Language and history on Kilimanjaro, the Taita Hills, and the Pare Mountains, with 24 tables and 3 maps. Hamburg: Helmut Buske. Nurse, Derek 1982. Segeju and Daisũ: A case study of evidence from oral traditions and comparative linguistics. History in Africa 9:175-208. Nurse, Derek 1986. Reconstruction of Dahalo history through evidence from loanwords. Sprache und Geschichte in Afrika (SUGIA) 7: 267-305. Nurse, Derek 1994. South meets north: Ilwana = Bantu+Cushitic on Kenya's Tana River. In Bakker and Mous (eds.) Mixed languages: 15 case studies in language intertwining, pp. 213-222. Amsterdam: IFOTT. Nurse, Derek 2000. Inheritance, contact, and change in two East African languages. Cologne: Rüdiger Köppe. Nurse, Derek and Thomas Hinnebush 1993. Swahili and Sabaki: A linguistic history. (University of California Publications, Linguistics, 121.) Berkeley/Los Angeles: University of California Press. Nurse, Derek and Gérard Philippson 1975. The north-eastern Bantu languages of Tan­ zania and Kenya: A classification. Kiswahili 45(2): 1-28. O'Barr, William 1973. The Pare of Tanzania. In Mark A. Tessler, William O'Barr and David Psain (eds.) Tradition and identity in changing Africa, pp. 17-100. New York: Harper & Row. Odden, David 1982. Tonal phenomena in Kishambaa. Studies in African Linguistics 13: 177-208. Odden, David 1986. Three dialects of Kipare. In Gerrit J. Dimmendaal (ed.) Current approaches to African linguistics, vol. 3, pp. 257-280. (Publications in African Languages and Linguistics, 6.) Dordrecht: Foris Owens, Jonathan 1980. Observations on tone in the Booran dialect of Oromo. African Language Studies 17:141-196. Owens, Jonathan 1991/92. Review of Thomason and Kaufman (1988). Sprache und Geschichte in Afrika (SUGIA) 12/13:401 -409. Philippson, Gérard 1984. "Gens des bananeraies" : Contribution linguistique à l'histoire culturelle des chaga du Kilimanjaro. ('Cahier', 16.) Paris: Editions Re­ cherche sur les Civilisations. Philippson, Gérard n.d. Word lists Chasu, Saghala. Ms. Philippson, Gérard and Derek Nurse 2000. Gweno, a little known Bantu language of Northern Tanzania. In Kulikoyela Kahigi, Yared Kihore and Maarten Mous (eds.) Lugha za Tanzania / Languages of Tanzania: Studies dedicated to the memory of Prof Clement Maganga, pp. 233-284. Leiden: CNWS. Reh, Mechthild 1995. Möglichkeiten und Grenzen der Rekonstruirbarkeit von Soziohistorie aus Sprachdaten; am Beispiel der Kadugli-Krongo (NubaBerge, Sudan) Ms. cited in Kießling (1998). Rekdal, Ole Bjørn 1998. When hypothesis becomes myth: The Iraqi origin of the Iraqw. Ethnology 37: 17-38.

228

REFERENCES

Rikli, Martin 1929. Bei den Wambugu im Usambaragebirge. Übersee- und Kolonial­ zeitung; Der Kolionaldeutsche 9:153-154. Roehl, Karl 1911. Versuch einer systematischen Grammatik der Schambalasprache, (Abhandlungen des Hamburgischen Kolonialinstituts, 2 Reihe B: Völkerkunde, Kulturgeschichte und Sprachen, 2.) Hamburg: L. Friederichsen. Ross, Malcolm D. 1991. Refining Guy's sociolinguistic types of language change. Diachronica 8:119-129. Ross, Malcolm D. 1996. Contact-induced change and the comparative method: Cases from Papua New Guinea. In Mark Durie and Malcolm Ross (eds.) The comparative method reviewed: Regularity and irregularity in language change, pp. 180-217. New York: Oxford University Press. Rottland, Franz 1982. Die siidnilotischen Sprachen. Beschreibung, Vergleichung und Rekonstruktion. Berlin: Dietrich Reimer. Sacleux, Ch. 1939. Dictionnaire Swahili - Français. Paris: Institut d'Ethnologie. Sasse, Hans-Jürgen 1979. The consonant phonemes of Proto-East-Cushitic (PEC): A first approximation. (Afroasiatic Linguistics, 7.1). Malibu : Undena Publications. Sasse, Hans-Jürgen 1992. Theory of language death. In Brenzinger (ed.) Tanguage death: Factual and theoretical explorations with special reference to East Africa, pp. 7-30. Berlin: Mouton-De Gruyter. Sommer, Gabriele and Rainer Vossen 1993. Dialects, sectiolects, or simply lects? The Maa language in time perspective. In Thomas Spear and Richard Waller (eds.) Be­ ing Maasai: Ethnicity and Identity in East Africa, pp 25-37. London: James Curry. Smith, Norval 1995. An annotated list of creoles, pidgins, and mixed languages. In Jacques Arends, Pieter Muysken and Norval Smith (eds.) Pidgins and Creoles: An Introduction, pp. 331-374. Amsterdam: John Benjamins. Storch. 1895. Sitten, Gebräuche und Rechtspflege bei den Bewohnern Usambaras und Pares. Mittheilungen von Forschungsreisenden und Gelehrten aus den Deutschen Schutzgebieten 8:310-331.. Stroomer, Harry 1987. A comparative study of three Southern Oromo dialects in Kenya (Kuschitische Sprachstudien, 6.) Hamburg: Helmut Buske. Stroomer, Harry 1995. A grammar of Boraana Oromo (Kenya). (Kuschitische Sprachstudien, 11.) Cologne: Rüdiger Köppe. Tate, H.R. 1904. Notes on the Kikuyu and Kamba tribes of British East Africa. Jour­ nal of the Anthropological Institute of Great Britain and Ireland 34:130-148. Thomason, Sarah. G. 1983. Genetic relationship and the case of Ma'a (Mbugu). Stud­ ies in African linguistics 14:195-231. Thomason, Sarah G. 1995. Language mixture: Ordinary processes, extraordinary re­ sults. In Carmen Silva-Corvalán (ed.) Spanish in four continents: Studies in lan­ guage contact and bilingualism, pp. 15-23. Washington DC: Georgetown Univer­ sity Press. Thomason, Sarah G. 1997. Ma'a (Mbugu). In Sarah G. Thomason (ed.) Contact Lan­ guages: A wider perpective, pp. 469-487. Amsterdam: John Benjamins.

REFERENCES

229

Thomason, Sarah G. 2001. Language Contact: An introduction. Edinburgh: Edinburgh University Press. Thomason, Sarah G. and Terrence Kaufman 1988. Language contact, creolization, and genetic linguistics. Berkeley: University of California Press. Torry, W.Í. 1973. Subsistence ecology among the Gabbra: Nomads of the Kenya / Ethiopia frontier. PhD dissertation, Columbia University. Tosco, Mauro 1989. The classification of Dahalo: Another perspective Paper pre­ sented at the Second International Symposium on Cushitic and Omotic languages, Turin, November 16-18, 1989. Tosco, Mauro 1991. A grammatical sketch of Dahalo including texts and a glossary. (Kuschitische Sprachstudien, 8.) Hamburg: Helmut Buske. Tosco, Mauro 1992. Dahalo: An endangered language. In Mathias Brenzinger (ed.) Language death: Factual and theoretical explorations with special reference to East Africa, pp. 137-155. Berlin: Mouton-De Gruyter. Tucker, Archibald. N. and Margaret. A. Bryan 1974. The 'Mbugu' Anomaly. Bulletin of the School of Oriental and African Studies 37:188-207. Van Driem, Georg. 2001. Languages of the Himalayas: An ethnolinguistic handbook of the greater Himalyan region: containing an introduction to the symbiotic theory of language. Leiden: Brill. Van Minde, Don 1997. Malayu Ambong: Phonology, morphology, syntax. (CNWS publications, 61.) Leiden: CNWS. Walsh, Martin 1990. The Degere: Forgotten hunter-gatherers of the East African Coast. Cambridge Anthropology 14(3):68-81. Walsh, Martin 1992/93. The Vuna and the Degere: Remnants and Outcasts among the Duruma and Digo of Kenya and Tanzania. Bulletin of the International Committee on Urgent Anthropological and Ethnological Research 34/35:133-147. Weimers, William E. 1966. African Language Structures. Berkeley and Los Angeles: University of California Press. Whiteley, William H. 1960. Linguistic hybrids. African Studies 19:95-97. Williamson, John 1943. Dabida numerals. African Studies 2:215-216. Winans, Edgar V. 1962. Shambala: The constitution o f a traditional state. London: Routledge and Kegan Paul. Winter, Jürgen-Christoph 1979. Language shift among the Aasáx, a hunter-gatherer tribe in Tanzania: A historical and sociolinguistic case-study. Sprache und Geschichte in Afrika (SUGIA) 1:175-204. Wolff, Hans 1965. Review of Hans G. Mukarovsky 'Die Grundlagen des Ful und das Mauretanische' (1963). American Anthropologist 67:593-594. Woodward, H.W. 1913/14. Kitaita or Kisighau as spoken on the Shambala hills above Bwiti. Zeitschrift für Kolonial sprachen 4:91-117. Wray, J.A. 1894. An elementary introduction to the Taita language. London. Zaborski, Andrzej 1984. Review of Ehret (1980). Afrika und Übersee 67:142-144.

APPENDIX: MBUGU - ENGLISH ETYMOLOGICAL LEXICON

The Mbugu wordlist of over 2300 lemmas is organised as follows: The first (sort) field is the lemma (in italics); in the alphabetical order ŋ is placed after n, and ' before a. The next field is the word category (between brackets). The following abbreviations are used: (adv) adverb (aj) adjective (dem) demonstrative (ideo) ideophone (inv) invariable (n) noun (1/2) noun class 1, plural 2 (num) numeral (poss) possessive (pro) pronoun (ques) question word (v) verb (v ap) verb with applied extension (v cs) verb causative extension (v ps) verb with passive extension (v rc) verb with reciprocal extension (v sp) verb with separative extension (v st) verb with stative extension (v:ap) verb with lexicalised applied extension, etc. The next field is the meaning. Here I make no principled distinction between different senses; all are separated by commas. The abbreviation n.pr.pop. (proper noun for peoples) indicates names of peoples (nations), and n.pr.pers. (proper noun for per­ sons) indicates personal names, with the addition fem for feminine names and masc, for masculine names. The next field gives an indication of the language [between square brackets], and the equivalent of the lemma in the parallel lexicon is given. The following abbre­ viations are used: IMb for Inner Mbugu (Ma'á), IMb-Bum for the Bumbuli dialect of

232

APPENDIX

IMb, IMb-Ra for the Rangwi dialect of IMb; NMb for Normal Mbugu; and I+NMb for Inner and Normal Mbugu. The next fields represent variants of the lemma which are preceded by "also" and related lemmas which are preceded by "cf.". For derived lemmas the base is given after "Der.'' Occasionally an example of usage is given after "E.g.". The last field gives the etymological source after "